1 2020-02-18 Martin Liska <mliska@suse.cz>
4 * cgraph.c (cgraph_node::verify_node): Verify MALLOC attribute
5 and return type of functions.
6 * ipa-param-manipulation.c (ipa_param_adjustments::adjust_decl):
7 Drop MALLOC attribute for void functions.
8 * ipa-pure-const.c (funct_state_summary_t::duplicate): Drop
9 malloc_state for a new VOID clone.
11 2020-02-18 Martin Liska <mliska@suse.cz>
14 * common.opt: Add -fprofile-reproducibility.
15 * doc/invoke.texi: Document it.
16 * value-prof.c (dump_histogram_value):
17 Document and support behavior for counters[0]
18 being a negative value.
19 (get_nth_most_common_value): Handle negative
20 counters[0] in respect to flag_profile_reproducible.
22 2020-02-18 Jakub Jelinek <jakub@redhat.com>
25 * cgraph.c (verify_speculative_call): Use speculative_id instead of
26 speculative_uid in messages. Remove trailing whitespace from error
27 message. Use num_speculative_call_targets instead of
28 num_speculative_targets in a message.
29 (cgraph_node::verify_node): Use call_stmt instead of cal_stmt in
30 edge messages and stmt instead of cal_stmt in reference message.
32 PR tree-optimization/93780
33 * tree-ssa.c (non_rewritable_lvalue_p): Check valid_vector_subparts_p
34 before calling build_vector_type.
35 (execute_update_addresses_taken): Likewise.
38 * params.opt (-param=ipa-max-switch-predicate-bounds=): Fix help
39 typo, functoin -> function.
40 * tree.c (free_lang_data_in_decl): Fix comment typo,
42 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
44 2020-02-17 David Malcolm <dmalcolm@redhat.com>
46 * diagnostic.c (print_any_cwe): Don't call get_cwe_url if URLs
48 (print_option_information): Don't call get_option_url if URLs
51 2020-02-17 Alexandre Oliva <oliva@adacore.com>
53 * tree-emutls.c (new_emutls_decl, emutls_common_1): Complete
54 handling of register_common-less targets.
56 2020-02-17 Martin Liska <mliska@suse.cz>
59 * ipa-devirt.c (odr_types_equivalent_p): Fix grammar.
61 2020-02-17 Martin Liska <mliska@suse.cz>
64 * config/rs6000/rs6000.c (rs6000_option_override_internal):
67 2020-02-17 Martin Liska <mliska@suse.cz>
70 * config/rx/elf.opt: Fix typo.
72 2020-02-17 Richard Biener <rguenther@suse.de>
75 * opts-global.c (print_ignored_options): Use inform and
78 2020-02-17 Jiufu Guo <guojiufu@linux.ibm.com>
81 * config/rs6000/rs6000.md (untyped_call): Add emit_clobber.
83 2020-02-16 Uroš Bizjak <ubizjak@gmail.com>
86 * config/i386/i386.md (atan2xf3): Swap operands 1 and 2.
87 (atan2<mode>3): Update operand order in the call to gen_atan2xf3.
89 2020-02-15 Jason Merrill <jason@redhat.com>
91 * doc/invoke.texi (C Dialect Options): Add -std=c++20.
93 2020-02-15 Jakub Jelinek <jakub@redhat.com>
95 PR tree-optimization/93744
96 * match.pd (((m1 >/</>=/<= m2) * d -> (m1 >/</>=/<= m2) ? d : 0,
97 A - ((A - B) & -(C cmp D)) -> (C cmp D) ? B : A,
98 A + ((B - A) & -(C cmp D)) -> (C cmp D) ? B : A): For GENERIC, make
99 sure @2 in the first and @1 in the other patterns has no side-effects.
101 2020-02-15 David Malcolm <dmalcolm@redhat.com>
102 Bernd Edlinger <bernd.edlinger@hotmail.de>
106 * config.in (DIAGNOSTICS_URLS_DEFAULT): New define.
107 * configure.ac (--with-diagnostics-urls): New configuration
108 option, based on --with-diagnostics-color.
109 (DIAGNOSTICS_URLS_DEFAULT): New define.
110 * config.h: Regenerate.
111 * configure: Regenerate.
112 * diagnostic.c (diagnostic_urls_init): Handle -1 for
113 DIAGNOSTICS_URLS_DEFAULT from configure-time
114 --with-diagnostics-urls=auto-if-env by querying for a GCC_URLS
115 and TERM_URLS environment variable.
116 * diagnostic-url.h (diagnostic_url_format): New enum type.
117 (diagnostic_urls_enabled_p): rename to...
118 (determine_url_format): ... this, and change return type.
119 * diagnostic-color.c (parse_env_vars_for_urls): New helper function.
120 (auto_enable_urls): Disable URLs on xfce4-terminal, gnome-terminal,
121 the linux console, and mingw.
122 (diagnostic_urls_enabled_p): rename to...
123 (determine_url_format): ... this, and adjust.
124 * pretty-print.h (pretty_printer::show_urls): rename to...
125 (pretty_printer::url_format): ... this, and change to enum.
126 * pretty-print.c (pretty_printer::pretty_printer,
127 pp_begin_url, pp_end_url, test_urls): Adjust.
128 * doc/install.texi (--with-diagnostics-urls): Document the new
129 configuration option.
130 (--with-diagnostics-color): Document the existing interaction
131 with GCC_COLORS better.
132 * doc/invoke.texi (-fdiagnostics-urls): Add GCC_URLS and TERM_URLS
133 vindex reference. Update description of defaults based on the above.
134 (-fdiagnostics-color): Update description of how -fdiagnostics-color
135 interacts with GCC_COLORS.
137 2020-02-14 Eric Botcazou <ebotcazou@adacore.com>
140 * config/sparc/sparc.c (eligible_for_call_delay): Test HAVE_GNU_LD in
141 conjunction with TARGET_GNU_TLS in early return.
143 2020-02-14 Alexander Monakov <amonakov@ispras.ru>
145 * rtlanal.c (rtx_cost): Handle a SET up front. Avoid division if
146 the mode is not wider than UNITS_PER_WORD.
148 2020-02-14 Martin Jambor <mjambor@suse.cz>
150 PR tree-optimization/93516
151 * tree-sra.c (propagate_subaccesses_from_rhs): Do not create
152 access of the same type as the parent.
153 (propagate_subaccesses_from_lhs): Likewise.
155 2020-02-14 Hongtao Liu <hongtao.liu@intel.com>
158 * config/i386/avx512vbmi2intrin.h
159 (_mm512_shrdi_epi16, _mm512_mask_shrdi_epi16,
160 _mm512_maskz_shrdi_epi16, _mm512_shrdi_epi32,
161 _mm512_mask_shrdi_epi32, _mm512_maskz_shrdi_epi32,
162 _m512_shrdi_epi64, _m512_mask_shrdi_epi64,
163 _m512_maskz_shrdi_epi64, _mm512_shldi_epi16,
164 _mm512_mask_shldi_epi16, _mm512_maskz_shldi_epi16,
165 _mm512_shldi_epi32, _mm512_mask_shldi_epi32,
166 _mm512_maskz_shldi_epi32, _mm512_shldi_epi64,
167 _mm512_mask_shldi_epi64, _mm512_maskz_shldi_epi64): Fix typo
168 of lacking a closing parenthesis.
169 * config/i386/avx512vbmi2vlintrin.h
170 (_mm256_shrdi_epi16, _mm256_mask_shrdi_epi16,
171 _mm256_maskz_shrdi_epi16, _mm256_shrdi_epi32,
172 _mm256_mask_shrdi_epi32, _mm256_maskz_shrdi_epi32,
173 _m256_shrdi_epi64, _m256_mask_shrdi_epi64,
174 _m256_maskz_shrdi_epi64, _mm256_shldi_epi16,
175 _mm256_mask_shldi_epi16, _mm256_maskz_shldi_epi16,
176 _mm256_shldi_epi32, _mm256_mask_shldi_epi32,
177 _mm256_maskz_shldi_epi32, _mm256_shldi_epi64,
178 _mm256_mask_shldi_epi64, _mm256_maskz_shldi_epi64,
179 _mm_shrdi_epi16, _mm_mask_shrdi_epi16,
180 _mm_maskz_shrdi_epi16, _mm_shrdi_epi32,
181 _mm_mask_shrdi_epi32, _mm_maskz_shrdi_epi32,
182 _mm_shrdi_epi64, _mm_mask_shrdi_epi64,
183 _m_maskz_shrdi_epi64, _mm_shldi_epi16,
184 _mm_mask_shldi_epi16, _mm_maskz_shldi_epi16,
185 _mm_shldi_epi32, _mm_mask_shldi_epi32,
186 _mm_maskz_shldi_epi32, _mm_shldi_epi64,
187 _mm_mask_shldi_epi64, _mm_maskz_shldi_epi64): Ditto.
189 2020-02-13 H.J. Lu <hongjiu.lu@intel.com>
192 * config/i386/i386.c (ix86_trampoline_init): Skip ENDBR32 at
193 the target function entry.
195 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
197 * common/config/arc/arc-common.c (arc_option_optimization_table):
198 Disable if-conversion step when optimized for size.
200 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
202 * config/arc/arc.c (arc_conditional_register_usage): R0-R3 and
203 R12-R15 are always in ARCOMPACT16_REGS register class.
204 * config/arc/arc.opt (mq-class): Deprecate.
205 * config/arc/constraint.md ("q"): Remove dependency on mq-class
207 * doc/invoke.texi (mq-class): Update text.
208 * common/config/arc/arc-common.c (arc_option_optimization_table):
211 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
213 * config/arc/arc.c (arc_insn_cost): New function.
214 (TARGET_INSN_COST): Define.
215 * config/arc/arc.md (cost): New attribute.
216 (add_n): Use arc_nonmemory_operand.
217 (ashlsi3_insn): Likewise, also update constraints.
218 (ashrsi3_insn): Likewise.
220 (add_shift): Likewise.
221 * config/arc/predicates.md (arc_nonmemory_operand): New predicate.
223 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
225 * config/arc/arc.md (mulsidi_600): Correctly select mlo/mhi
227 (umulsidi_600): Likewise.
229 2020-02-13 Jakub Jelinek <jakub@redhat.com>
232 * config/i386/avx512bitalgintrin.h (_mm512_mask_popcnt_epi8,
233 _mm512_mask_popcnt_epi16, _mm256_mask_popcnt_epi8,
234 _mm256_mask_popcnt_epi16, _mm_mask_popcnt_epi8,
235 _mm_mask_popcnt_epi16): Rename __B argument to __A and __A to __W,
236 pass __A to the builtin followed by __W instead of __A followed by
238 * config/i386/avx512vpopcntdqintrin.h (_mm512_mask_popcnt_epi32,
239 _mm512_mask_popcnt_epi64): Likewise.
240 * config/i386/avx512vpopcntdqvlintrin.h (_mm_mask_popcnt_epi32,
241 _mm256_mask_popcnt_epi32, _mm_mask_popcnt_epi64,
242 _mm256_mask_popcnt_epi64): Likewise.
244 PR tree-optimization/93582
245 * fold-const.h (shift_bytes_in_array_left,
246 shift_bytes_in_array_right): Declare.
247 * fold-const.c (shift_bytes_in_array_left,
248 shift_bytes_in_array_right): New function, moved from
249 gimple-ssa-store-merging.c, no longer static.
250 * gimple-ssa-store-merging.c (shift_bytes_in_array): Move
251 to gimple-ssa-store-merging.c and rename to shift_bytes_in_array_left.
252 (shift_bytes_in_array_right): Move to gimple-ssa-store-merging.c.
253 (encode_tree_to_bitpos): Use shift_bytes_in_array_left instead of
254 shift_bytes_in_array.
255 (verify_shift_bytes_in_array): Rename to ...
256 (verify_shift_bytes_in_array_left): ... this. Use
257 shift_bytes_in_array_left instead of shift_bytes_in_array.
258 (store_merging_c_tests): Call verify_shift_bytes_in_array_left
259 instead of verify_shift_bytes_in_array.
260 * tree-ssa-sccvn.c (vn_reference_lookup_3): For native_encode_expr
261 / native_interpret_expr where the store covers all needed bits,
262 punt on PDP-endian, otherwise allow all involved offsets and sizes
263 not to be byte-aligned.
266 * config/i386/sse.md (k<code><mode>): Drop mode from last operand and
267 use const_0_to_255_operand predicate instead of immediate_operand.
268 (avx512dq_fpclass<mode><mask_scalar_merge_name>,
269 avx512dq_vmfpclass<mode><mask_scalar_merge_name>,
270 vgf2p8affineinvqb_<mode><mask_name>,
271 vgf2p8affineqb_<mode><mask_name>): Drop mode from
272 const_0_to_255_operand predicated operands.
274 2020-02-12 Jeff Law <law@redhat.com>
276 * config/h8300/h8300.md (comparison shortening peepholes): Use
277 a mode iterator to merge the HImode and SImode peepholes.
279 2020-02-12 Jakub Jelinek <jakub@redhat.com>
282 * real.c (is_even): Make static. Function comment fix.
283 (is_halfway_below): Make static, don't assert R is not inf/nan,
284 instead return false for those. Small formatting fixes.
286 2020-02-12 Martin Sebor <msebor@redhat.com>
289 * tree-ssa-strlen.c (handle_builtin_stxncpy): Rename...
290 (handle_builtin_stxncpy_strncat): ...to this. Change first argument.
291 Issue only -Wstringop-overflow strncat, never -Wstringop-truncation.
292 (strlen_check_and_optimize_call): Adjust callee name.
294 2020-02-12 Jeff Law <law@redhat.com>
296 * config/h8300/h8300.md (comparison shortening peepholes): Drop
297 (and (xor)) variant. Combine other two into single peephole.
299 2020-02-12 Wilco Dijkstra <wdijkstr@arm.com>
301 PR rtl-optimization/93565
302 * config/aarch64/aarch64.c (aarch64_rtx_costs): Add CTZ costs.
304 2020-02-12 Wilco Dijkstra <wdijkstr@arm.com>
306 * config/aarch64/aarch64-simd.md
307 (aarch64_zero_extend<GPI:mode>_reduc_plus_<VDQV_E:mode>): New pattern.
308 * config/aarch64/aarch64.md (popcount<mode>2): Use it instead of
309 generating separate ADDV and zero_extend patterns.
310 * config/aarch64/iterators.md (VDQV_E): New iterator.
312 2020-02-12 Jeff Law <law@redhat.com>
314 * config/h8300/h8300.md (cpymemsi, movmd): Remove dead patterns,
315 expanders, splits, etc.
316 (movmd_internal_<mode>, movmd splitter, movstr, movsd): Likewise.
317 (stpcpy_internal_<mode>, stpcpy splitter): Likewise.
318 (peepholes to convert QI/HI mode pushes to SI mode pushes): Likewise.
319 * config/h8300/h8300.c (h8300_swap_into_er6): Remove unused function.
320 (h8300_swap_out_of_er6, h8sx_emit_movmd): Likewise
321 * config/h8300/h8300-protos.h (h8300_swap_into_er6): Remove unused
323 (h8300_swap_out_of_er6, h8sx_emit_movmd): Likewise.
325 2020-02-12 Jakub Jelinek <jakub@redhat.com>
328 * config/i386/sse.md (VI48F_256_DQ): New mode iterator.
329 (avx512vl_vextractf128<mode>): Use it instead of VI48F_256. Remove
330 TARGET_AVX512DQ from condition.
331 (vec_extract_lo_<mode><mask_name>): Use <mask_avx512dq_condition>
332 instead of <mask_mode512bit_condition> in condition. If
333 TARGET_AVX512DQ is false, emit vextract*64x4 instead of
335 (vec_extract_lo_<mode><mask_name>): Drop <mask_avx512dq_condition>
338 2020-02-12 Kewen Lin <linkw@gcc.gnu.org>
341 * ira.c (combine_and_move_insns): Skip multiple_sets def_insn.
343 2020-02-12 Segher Boessenkool <segher@kernel.crashing.org>
345 * config/rs6000/rs6000.c (rs6000_debug_print_mode): Don't use sizeof
346 where strlen is more legible.
347 (rs6000_builtin_vectorized_libmass): Ditto.
348 (rs6000_print_options_internal): Ditto.
350 2020-02-11 Martin Sebor <msebor@redhat.com>
352 PR tree-optimization/93683
353 * tree-ssa-alias.c (stmt_kills_ref_p): Avoid using LHS when not set.
355 2020-02-11 Michael Meissner <meissner@linux.ibm.com>
357 * config/rs6000/predicates.md (cint34_operand): Rename the
358 -mprefixed-addr option to be -mprefixed.
359 * config/rs6000/rs6000-cpus.def (ISA_FUTURE_MASKS_SERVER): Rename
360 the -mprefixed-addr option to be -mprefixed.
361 (OTHER_FUTURE_MASKS): Likewise.
362 (POWERPC_MASKS): Likewise.
363 * config/rs6000/rs6000.c (rs6000_option_override_internal): Rename
364 the -mprefixed-addr option to be -mprefixed. Change error
365 messages to refer to -mprefixed.
366 (num_insns_constant_gpr): Rename the -mprefixed-addr option to be
368 (rs6000_legitimate_offset_address_p): Likewise.
369 (rs6000_mode_dependent_address): Likewise.
370 (rs6000_opt_masks): Change the spelling of "-mprefixed-addr" to be
371 "-mprefixed" for target attributes and pragmas.
372 (address_to_insn_form): Rename the -mprefixed-addr option to be
374 (rs6000_adjust_insn_length): Likewise.
375 * config/rs6000/rs6000.h (FINAL_PRESCAN_INSN): Rename the
376 -mprefixed-addr option to be -mprefixed.
377 (ASM_OUTPUT_OPCODE): Likewise.
378 * config/rs6000/rs6000.md (prefixed insn attribute): Rename the
379 -mprefixed-addr option to be -mprefixed.
380 * config/rs6000/rs6000.opt (-mprefixed): Rename the
381 -mprefixed-addr option to be prefixed. Change the option from
382 being undocumented to being documented.
383 * doc/invoke.texi (RS/6000 and PowerPC Options): Document the
384 -mprefixed option. Update the -mpcrel documentation to mention
387 2020-02-11 Hans-Peter Nilsson <hp@axis.com>
389 * ira-conflicts.c (print_hard_reg_set): Correct output for sets
390 including FIRST_PSEUDO_REGISTER - 1.
391 * ira-color.c (print_hard_reg_set): Ditto.
393 2020-02-11 Stam Markianos-Wright <stam.markianos-wright@arm.com>
395 * config/arm/arm-builtins.c (enum arm_type_qualifiers):
396 (USTERNOP_QUALIFIERS): New define.
397 (USMAC_LANE_QUADTUP_QUALIFIERS): New define.
398 (SUMAC_LANE_QUADTUP_QUALIFIERS): New define.
399 (arm_expand_builtin_args): Add case ARG_BUILTIN_LANE_QUADTUP_INDEX.
400 (arm_expand_builtin_1): Add qualifier_lane_quadtup_index.
401 * config/arm/arm_neon.h (vusdot_s32): New.
402 (vusdot_lane_s32): New.
403 (vusdotq_lane_s32): New.
404 (vsudot_lane_s32): New.
405 (vsudotq_lane_s32): New.
406 * config/arm/arm_neon_builtins.def (usdot, usdot_lane,sudot_lane): New.
407 * config/arm/iterators.md (DOTPROD_I8MM): New.
408 (sup, opsuffix): Add <us/su>.
409 * config/arm/neon.md (neon_usdot, <us/su>dot_lane: New.
410 * config/arm/unspecs.md (UNSPEC_DOT_US, UNSPEC_DOT_SU): New.
412 2020-02-11 Richard Biener <rguenther@suse.de>
414 PR tree-optimization/93661
415 PR tree-optimization/93662
416 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly guard
418 * tree-sra.c (get_access_for_expr): Likewise.
420 2020-02-10 Jakub Jelinek <jakub@redhat.com>
423 * config/i386/sse.md (VI_256_AVX2): New mode iterator.
424 (vcond_mask_<mode><sseintvecmodelower>): Use it instead of VI_256.
425 Change condition from TARGET_AVX2 to TARGET_AVX.
427 2020-02-10 Iain Sandoe <iain@sandoe.co.uk>
430 * config/darwin-c.c (darwin_cfstring_ref_p): Fix up last
433 2020-02-10 Hans-Peter Nilsson <hp@axis.com>
435 Try to generate zero-based comparisons.
436 * config/cris/cris.c (cris_reduce_compare): New function.
437 * config/cris/cris-protos.h (cris_reduce_compare): Add prototype.
438 * config/cris/cris.md ("cbranch<mode>4", "cbranchdi4", "cstoredi4")
439 (cstore<mode>4"): Apply cris_reduce_compare in expanders.
441 2020-02-10 Richard Earnshaw <rearnsha@arm.com>
444 * config/arm/arm.md (movsi_compare0): Allow SP as a source register
445 in Thumb state and also as a destination in Arm state. Add T16
448 2020-02-10 Hans-Peter Nilsson <hp@axis.com>
450 * md.texi (Define Subst): Match closing paren in example.
452 2020-02-10 Jakub Jelinek <jakub@redhat.com>
456 * config/i386/i386.c (x86_64_elf_section_type_flags): Fix up last
457 arguments of strncmp.
459 2020-02-10 Feng Xue <fxue@os.amperecomputing.com>
462 * ipa-cp.c (ipcp_lattice::add_value): Add source with same call edge
463 but different source value.
464 (adjust_callers_for_value_intersection): New function.
465 (gather_edges_for_value): Adjust order of callers to let a
466 non-self-recursive caller be the first element.
467 (self_recursive_pass_through_p): Add a new parameter "simple", and
468 check generalized self-recursive pass-through jump function.
469 (self_recursive_agg_pass_through_p): Likewise.
470 (find_more_scalar_values_for_callers_subset): Compute value from
471 pass-through jump function for self-recursive.
472 (intersect_with_plats): Cleanup previous implementation code for value
473 itersection with self-recursive call edge.
474 (intersect_with_agg_replacements): Likewise.
475 (intersect_aggregates_with_edge): Deduce value from pass-through jump
476 function for self-recursive call edge. Cleanup previous implementation
477 code for value intersection with self-recursive call edge.
478 (decide_whether_version_node): Remove dead callers and adjust order
479 to let a non-self-recursive caller be the first element.
481 2020-02-09 Uroš Bizjak <ubizjak@gmail.com>
483 * recog.c: Move pass_split_before_sched2 code in front of
484 pass_split_before_regstack.
485 (pass_data_split_before_sched2): Rename pass to split3 from split4.
486 (pass_data_split_before_regstack): Rename pass to split4 from split3.
487 (rest_of_handle_split_before_sched2): Remove.
488 (pass_split_before_sched2::execute): Unconditionally call
490 (enable_split_before_sched2): New function.
491 (pass_split_before_sched2::gate): Use enable_split_before_sched2.
492 (pass_split_before_regstack::gate): Ditto.
493 * config/nds32/nds32.c (nds32_split_double_word_load_store_p):
494 Update name check for renamed split4 pass.
495 * config/sh/sh.c (register_sh_passes): Update pass insertion
496 point for renamed split4 pass.
498 2020-02-09 Jakub Jelinek <jakub@redhat.com>
500 * gimplify.c (gimplify_adjust_omp_clauses_1): Promote
501 DECL_IN_CONSTANT_POOL variables into "omp declare target" to avoid
502 copying them around between host and target.
504 2020-02-08 Andrew Pinski <apinski@marvell.com>
507 * config/aarch64/aarch64-simd.md (movmisalign<mode>): Check
508 STRICT_ALIGNMENT also.
510 2020-02-08 Jim Wilson <jimw@sifive.com>
513 * config/riscv/riscv.h (HARD_REGNO_CALLER_SAVE_MODE): Define.
515 2020-02-08 Uroš Bizjak <ubizjak@gmail.com>
516 Jakub Jelinek <jakub@redhat.com>
519 * config/i386/i386.h (CALL_USED_REGISTERS): Make
520 xmm16-xmm31 call-used even in 64-bit ms-abi.
522 2020-02-07 Dennis Zhang <dennis.zhang@arm.com>
524 * config/aarch64/aarch64-simd-builtins.def (simd_smmla): New entry.
525 (simd_ummla, simd_usmmla): Likewise.
526 * config/aarch64/aarch64-simd.md (aarch64_simd_<sur>mmlav16qi): New.
527 * config/aarch64/arm_neon.h (vmmlaq_s32, vmmlaq_u32): New.
530 2020-02-07 Richard Biener <rguenther@suse.de>
533 * tree-inline.c (fold_marked_statements): Do a PRE walk,
534 skipping unreachable regions.
535 (optimize_inline_calls): Skip folding stmts when we didn't
538 2020-02-07 H.J. Lu <hongjiu.lu@intel.com>
541 * config/i386/i386.c (function_arg_ms_64): Add a type argument.
542 Don't return aggregates with only SFmode and DFmode in SSE
544 (ix86_function_arg): Pass arg.type to function_arg_ms_64.
546 2020-02-07 Jakub Jelinek <jakub@redhat.com>
549 * config/rs6000/rs6000-logue.c
550 (rs6000_emit_probe_stack_range_stack_clash): Always use gen_add3_insn,
551 if it fails, move rs into end_addr and retry. Add
552 REG_FRAME_RELATED_EXPR note whenever it returns more than one insn or
553 the insn pattern doesn't describe well what exactly happens to
557 * config/i386/predicates.md (avx_identity_operand): Remove.
558 * config/i386/sse.md (*avx_vec_concat<mode>_1): Remove.
559 (avx_<castmode><avxsizesuffix>_<castmode>,
560 avx512f_<castmode><avxsizesuffix>_256<castmode>): Change patterns to
561 a VEC_CONCAT of the operand and UNSPEC_CAST.
562 (avx512f_<castmode><avxsizesuffix>_<castmode>): Change pattern to
563 a VEC_CONCAT of VEC_CONCAT of the operand and UNSPEC_CAST with
567 * config/i386/i386.c (ix86_lea_outperforms): Make sure to clear
568 recog_data.insn if distance_non_agu_define changed it.
570 2020-02-06 Michael Meissner <meissner@linux.ibm.com>
573 * config/rs6000/rs6000.c (reg_to_non_prefixed): Before ISA 3.0
574 we only had X-FORM (reg+reg) addressing for vectors. Also before
575 ISA 3.0, we only had X-FORM addressing for scalars in the
576 traditional Altivec registers.
578 2020-02-06 <zhongyunde@huawei.com>
579 Vladimir Makarov <vmakarov@redhat.com>
581 PR rtl-optimization/93561
582 * lra-assigns.c (spill_for): Check that tested hard regno is not out of
585 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
587 * config/aarch64/aarch64.md (aarch64_movk<mode>): Add a type
590 2020-02-06 Segher Boessenkool <segher@kernel.crashing.org>
592 * config/rs6000/rs6000.c (rs6000_emit_set_long_const): Handle the case
593 where the low and the high 32 bits are equal to each other specially,
594 with an rldimi instruction.
596 2020-02-06 Mihail Ionescu <mihail.ionescu@arm.com>
598 * config/arm/arm-cpus.in: Set profile M for armv8.1-m.main.
600 2020-02-06 Mihail Ionescu <mihail.ionescu@arm.com>
602 * config/arm/arm-tables.opt: Regenerate.
604 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
607 * config/aarch64/aarch64-protos.h (aarch64_movk_shift): Declare.
608 * config/aarch64/aarch64.c (aarch64_movk_shift): New function.
609 * config/aarch64/aarch64.md (aarch64_movk<mode>): New pattern.
611 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
613 PR rtl-optimization/87763
614 * config/aarch64/aarch64.md (*ashiftsi_extvdi_bfiz): New pattern.
616 2020-02-06 Delia Burduv <delia.burduv@arm.com>
618 * config/aarch64/aarch64-simd-builtins.def
619 (bfmlaq): New built-in function.
620 (bfmlalb): New built-in function.
621 (bfmlalt): New built-in function.
622 (bfmlalb_lane): New built-in function.
623 (bfmlalt_lane): New built-in function.
624 * config/aarch64/aarch64-simd.md
625 (aarch64_bfmmlaqv4sf): New pattern.
626 (aarch64_bfmlal<bt>v4sf): New pattern.
627 (aarch64_bfmlal<bt>_lane<q>v4sf): New pattern.
628 * config/aarch64/arm_neon.h (vbfmmlaq_f32): New intrinsic.
629 (vbfmlalbq_f32): New intrinsic.
630 (vbfmlaltq_f32): New intrinsic.
631 (vbfmlalbq_lane_f32): New intrinsic.
632 (vbfmlaltq_lane_f32): New intrinsic.
633 (vbfmlalbq_laneq_f32): New intrinsic.
634 (vbfmlaltq_laneq_f32): New intrinsic.
635 * config/aarch64/iterators.md (BF_MLA): New int iterator.
636 (bt): New int attribute.
638 2020-02-06 Uroš Bizjak <ubizjak@gmail.com>
640 * config/i386/i386.md (*pushtf): Emit "#" instead of
641 calling gcc_unreachable in insn output.
644 (*pushsf_rex64): Ditto for alternatives other than 1.
645 (*pushsf): Ditto for alternatives other than 1.
647 2020-02-06 Martin Liska <mliska@suse.cz>
649 PR gcov-profile/91971
650 PR gcov-profile/93466
651 * coverage.c (coverage_init): Revert mangling of
652 path into filename. It can lead to huge filename length.
653 Creation of subfolders seem more natural.
655 2020-02-06 Stam Markianos-Wright <stam.markianos-wright@arm.com>
658 * config/arm/arm.c (arm_block_arith_comp_libfuncs_for_mode): New.
659 (arm_init_libfuncs): Add BFmode support to block spurious BF libfuncs.
660 Use arm_block_arith_comp_libfuncs_for_mode for HFmode.
662 2020-02-06 Jakub Jelinek <jakub@redhat.com>
665 * config/i386/predicates.md (avx_identity_operand): New predicate.
666 * config/i386/sse.md (*avx_vec_concat<mode>_1): New
667 define_insn_and_split.
670 * omp-low.c (use_pointer_for_field): For nested constructs, also
671 look for map clauses on target construct.
672 (scan_omp_1_stmt) <case GIMPLE_OMP_TARGET>: Bump temporarily
673 taskreg_nesting_level.
676 * gimplify.c (gimplify_scan_omp_clauses) <do_notice>: If adding
677 shared clause, call omp_notice_variable on outer context if any.
679 2020-02-05 Jason Merrill <jason@redhat.com>
682 * symtab.c (symtab_node::nonzero_address): A DECL_COMDAT decl has
683 non-zero address even if weak and not yet defined.
685 2020-02-05 Martin Sebor <msebor@redhat.com>
687 PR tree-optimization/92765
688 * gimple-fold.c (get_range_strlen_tree): Handle MEM_REF and PARM_DECL.
689 * tree-ssa-strlen.c (compute_string_length): Remove.
690 (determine_min_objsize): Remove.
691 (get_len_or_size): Add an argument. Call get_range_strlen_dynamic.
692 Avoid using type size as the upper bound on string length.
693 (handle_builtin_string_cmp): Add an argument. Adjust.
694 (strlen_check_and_optimize_call): Pass additional argument to
695 handle_builtin_string_cmp.
697 2020-02-05 Uroš Bizjak <ubizjak@gmail.com>
699 * config/i386/i386.md (*pushdi2_rex64 peephole2): Remove.
700 (*pushdi2_rex64 peephole2): Unconditionally split after
702 (*ashl<mode>3_doubleword): Ditto.
703 (*<shift_insn><mode>3_doubleword): Ditto.
705 2020-02-05 Michael Meissner <meissner@linux.ibm.com>
708 * config/rs6000/rs6000.c (get_vector_offset): Fix
710 2020-02-05 Andrew Stubbs <ams@codesourcery.com>
712 * config/gcn/t-gcn-hsa (MULTILIB_OPTIONS): Use / not space.
714 2020-02-05 David Malcolm <dmalcolm@redhat.com>
717 (Special Functions for Debugging the Analyzer): Update description
718 of __analyzer_dump_exploded_nodes.
720 2020-02-05 Jakub Jelinek <jakub@redhat.com>
723 * config/i386/i386-features.c (ix86_add_reg_usage_to_vzeroupper): Only
724 include sets and not clobbers in the vzeroupper pattern.
725 * config/i386/sse.md (*avx_vzeroupper): Require in insn condition that
726 the parallel has 17 (64-bit) or 9 (32-bit) elts.
727 (*avx_vzeroupper_1): New define_insn_and_split.
730 * recog.c (pass_split_after_reload::gate): For STACK_REGS targets,
731 don't run when !optimize.
732 (pass_split_before_regstack::gate): For STACK_REGS targets, run even
735 2020-02-05 Richard Biener <rguenther@suse.de>
738 * genmatch.c (dt_node::gen_kids_1): Emit number of argument
739 checks before matching calls.
741 2020-02-05 Jakub Jelinek <jakub@redhat.com>
743 * tree-ssa-alias.c (aliasing_matching_component_refs_p): Fix up
744 function comment typo.
747 * omp-simd-clone.c (expand_simd_clones): If simd_clone_mangle or
748 simd_clone_create failed when i == 0, adjust clone->nargs by
751 2020-02-05 Martin Liska <mliska@suse.cz>
754 * doc/invoke.texi: Document that one should
755 not combine ASLR and -fpch.
757 2020-02-04 Richard Biener <rguenther@suse.de>
759 PR tree-optimization/93538
760 * match.pd (addr EQ/NE ptr): Amend to handle &ptr->x EQ/NE ptr.
762 2020-02-04 Richard Biener <rguenther@suse.de>
764 PR tree-optimization/91123
765 * tree-ssa-sccvn.c (vn_walk_cb_data::finish): New method.
766 (vn_walk_cb_data::last_vuse): New member.
767 (vn_walk_cb_data::saved_operands): Likewsie.
768 (vn_walk_cb_data::~vn_walk_cb_data): Release saved_operands.
769 (vn_walk_cb_data::push_partial_def): Use finish.
770 (vn_reference_lookup_2): Update last_vuse and use finish if
771 we've saved operands.
772 (vn_reference_lookup_3): Use finish and update calls to
773 push_partial_defs everywhere. When translating through
774 memcpy or aggregate copies save off operands and alias-set.
775 (eliminate_dom_walker::eliminate_stmt): Restore VN_WALKREWRITE
776 operation for redundant store removal.
778 2020-02-04 Richard Biener <rguenther@suse.de>
780 PR tree-optimization/92819
781 * tree-ssa-forwprop.c (simplify_vector_constructor): Avoid
782 generating more stmts than before.
784 2020-02-04 Martin Liska <mliska@suse.cz>
786 * config/arm/arm.c (arm_gen_far_branch): Move the function
787 outside of selftests.
789 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
791 * config/rs6000/rs6000.c (adjust_vec_address_pcrel): New helper
792 function to adjust PC-relative vector addresses.
793 (rs6000_adjust_vec_address): Call adjust_vec_address_pcrel to
794 handle vectors with PC-relative addresses.
796 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
798 * config/rs6000/rs6000.c (reg_to_non_prefixed): Add forward
800 (hard_reg_and_mode_to_addr_mask): Delete.
801 (rs6000_adjust_vec_address): If the original vector address
802 was REG+REG or REG+OFFSET and the element is not zero, do the add
803 of the elements in the original address before adding the offset
804 for the vector element. Use address_to_insn_form to validate the
805 address using the register being loaded, rather than guessing
806 whether the address is a DS-FORM or DQ-FORM address.
808 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
810 * config/rs6000/rs6000.c (get_vector_offset): New helper function
811 to calculate the offset in memory from the start of a vector of a
812 particular element. Add code to keep the element number in
813 bounds if the element number is variable.
814 (rs6000_adjust_vec_address): Move calculation of offset of the
815 vector element to get_vector_offset.
816 (rs6000_split_vec_extract_var): Do not do the initial AND of
817 element here, move the code to get_vector_offset.
819 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
821 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add some
824 2020-02-03 Segher Boessenkool <segher@kernel.crashing.org>
826 * config/rs6000/constraints.md: Improve documentation.
828 2020-02-03 Richard Earnshaw <rearnsha@arm.com>
831 * config/arm/t-arm: ($(srcdir)/config/arm/arm-tune.md)
832 ($(srcdir)/config/arm/arm-tables.opt): Use move-if-change.
834 2020-02-03 Andrew Stubbs <ams@codesourcery.com>
836 * config.gcc: Remove "carrizo" support.
837 * config/gcn/gcn-opts.h (processor_type): Likewise.
838 * config/gcn/gcn.c (gcn_omp_device_kind_arch_isa): Likewise.
839 * config/gcn/gcn.opt (gpu_type): Likewise.
840 * config/gcn/t-omp-device: Likewise.
842 2020-02-03 Stam Markianos-Wright <stam.markianos-wright@arm.com>
845 * config/arm/arm-protos.h: New function arm_gen_far_branch prototype.
846 * config/arm/arm.c (arm_gen_far_branch): New function
848 * config/arm/arm.md: Update b<cond> for Thumb2 range checks.
850 2020-02-03 Julian Brown <julian@codesourcery.com>
851 Tobias Burnus <tobias@codesourcery.com>
853 * doc/invoke.texi: Update mention of OpenACC version to 2.6.
855 2020-02-03 Jakub Jelinek <jakub@redhat.com>
858 * config/s390/s390.md (popcounthi2_z196): Fix up expander to emit
859 valid RTL to sum up the lowest and second lowest bytes of the popcnt
862 2020-02-02 Vladimir Makarov <vmakarov@redhat.com>
864 PR rtl-optimization/91333
865 * ira-color.c (struct allocno_color_data): Add member
867 (init_allocno_threads): Set the member up.
868 (bucket_allocno_compare_func): Add compare hard reg
871 2020-01-31 Sandra Loosemore <sandra@codesourcery.com>
873 nios2: Support for GOT-relative DW_EH_PE_datarel encoding.
875 * configure.ac [nios2-*-*]: Check HAVE_AS_NIOS2_GOTOFF_RELOCATION.
876 * config.in: Regenerated.
877 * configure: Regenerated.
878 * config/nios2/nios2.h (ASM_PREFERRED_EH_DATA_FORMAT): Fix handling
879 for PIC when HAVE_AS_NIOS2_GOTOFF_RELOCATION.
880 (ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX): New.
882 2020-02-01 Andrew Burgess <andrew.burgess@embecosm.com>
884 * configure: Regenerate.
886 2020-01-31 Vladimir Makarov <vmakarov@redhat.com>
888 PR rtl-optimization/91333
889 * ira-color.c (bucket_allocno_compare_func): Move conflict hard
890 reg preferences comparison up.
892 2020-01-31 Richard Sandiford <richard.sandiford@arm.com>
894 * config/aarch64/aarch64.h (TARGET_SVE_BF16): New macro.
895 * config/aarch64/aarch64-sve-builtins-sve2.h (svcvtnt): Move to
896 aarch64-sve-builtins-base.h.
897 * config/aarch64/aarch64-sve-builtins-sve2.cc (svcvtnt): Move to
898 aarch64-sve-builtins-base.cc.
899 * config/aarch64/aarch64-sve-builtins-base.h (svbfdot, svbfdot_lane)
900 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
902 * config/aarch64/aarch64-sve-builtins-base.cc (svbfdot, svbfdot_lane)
903 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
904 (svcvtnt): New functions.
905 * config/aarch64/aarch64-sve-builtins-base.def (svbfdot, svbfdot_lane)
906 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
907 (svcvtnt): New functions.
908 (svcvt): Add a form that converts f32 to bf16.
909 * config/aarch64/aarch64-sve-builtins-shapes.h (ternary_bfloat)
910 (ternary_bfloat_lane, ternary_bfloat_lanex2, ternary_bfloat_opt_n):
912 * config/aarch64/aarch64-sve-builtins-shapes.cc (parse_element_type):
913 Treat B as bfloat16_t.
914 (ternary_bfloat_lane_base): New class.
915 (ternary_bfloat_def): Likewise.
916 (ternary_bfloat): New shape.
917 (ternary_bfloat_lane_def): New class.
918 (ternary_bfloat_lane): New shape.
919 (ternary_bfloat_lanex2_def): New class.
920 (ternary_bfloat_lanex2): New shape.
921 (ternary_bfloat_opt_n_def): New class.
922 (ternary_bfloat_opt_n): New shape.
923 * config/aarch64/aarch64-sve-builtins.cc (TYPES_cvt_bfloat): New macro.
924 * config/aarch64/aarch64-sve.md (@aarch64_sve_<sve_fp_op>vnx4sf)
925 (@aarch64_sve_<sve_fp_op>_lanevnx4sf): New patterns.
926 (@aarch64_sve_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>)
927 (@cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
928 (*cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
929 (@aarch64_sve_cvtnt<VNx8BF_ONLY:mode>): Likewise.
930 * config/aarch64/aarch64-sve2.md (@aarch64_sve2_cvtnt<mode>): Key
931 the pattern off the narrow mode instead of the wider one.
932 * config/aarch64/iterators.md (VNx8BF_ONLY): New mode iterator.
933 (UNSPEC_BFMLALB, UNSPEC_BFMLALT, UNSPEC_BFMMLA): New unspecs.
934 (sve_fp_op): Handle them.
935 (SVE_BFLOAT_TERNARY_LONG): New int itertor.
936 (SVE_BFLOAT_TERNARY_LONG_LANE): Likewise.
938 2020-01-31 Richard Sandiford <richard.sandiford@arm.com>
940 * config/aarch64/arm_sve.h: Include arm_bf16.h.
941 * config/aarch64/aarch64-modes.def (BF): Move definition before
942 VECTOR_MODES. Remove separate VECTOR_MODES for V4BF and V8BF.
943 (SVE_MODES): Handle BF modes.
944 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
946 (aarch64_full_sve_mode): Likewise.
947 * config/aarch64/iterators.md (SVE_STRUCT): Add VNx16BF, VNx24BF
949 (SVE_FULL, SVE_FULL_HSD, SVE_ALL): Add VNx8BF.
950 (Vetype, Vesize, Vctype, VEL, Vel, VEL_INT, V128, v128, vwcore)
951 (V_INT_EQUIV, v_int_equiv, V_FP_EQUIV, v_fp_equiv, vector_count)
952 (insn_length, VSINGLE, vsingle, VPRED, vpred, VDOUBLE): Handle the
954 * config/aarch64/aarch64-sve-builtins.h (TYPE_bfloat): New
956 * config/aarch64/aarch64-sve-builtins.cc (TYPES_all_arith): New macro.
957 (TYPES_all_data): Add bf16.
958 (TYPES_reinterpret1, TYPES_reinterpret): Likewise.
959 (register_tuple_type): Increase buffer size.
960 * config/aarch64/aarch64-sve-builtins.def (svbfloat16_t): New type.
961 (bf16): New type suffix.
962 * config/aarch64/aarch64-sve-builtins-base.def (svabd, svadd, svaddv)
963 (svcmpeq, svcmpge, svcmpgt, svcmple, svcmplt, svcmpne, svmad, svmax)
964 (svmaxv, svmin, svminv, svmla, svmls, svmsb, svmul, svsub, svsubr):
965 Change type from all_data to all_arith.
966 * config/aarch64/aarch64-sve-builtins-sve2.def (svaddp, svmaxp)
969 2020-01-31 Dennis Zhang <dennis.zhang@arm.com>
970 Matthew Malcomson <matthew.malcomson@arm.com>
971 Richard Sandiford <richard.sandiford@arm.com>
973 * doc/invoke.texi (f32mm): Document new AArch64 -march= extension.
974 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Define
975 __ARM_FEATURE_SVE_MATMUL_INT8, __ARM_FEATURE_SVE_MATMUL_FP32 and
976 __ARM_FEATURE_SVE_MATMUL_FP64 as appropriate. Don't define
977 __ARM_FEATURE_MATMUL_FP64.
978 * config/aarch64/aarch64-option-extensions.def (fp, simd, fp16)
979 (sve): Add AARCH64_FL_F32MM to the list of extensions that should
980 be disabled at the same time.
981 (f32mm): New extension.
982 * config/aarch64/aarch64.h (AARCH64_FL_F32MM): New macro.
983 (AARCH64_FL_F64MM): Bump to the next bit up.
984 (AARCH64_ISA_F32MM, TARGET_SVE_I8MM, TARGET_F32MM, TARGET_SVE_F32MM)
985 (TARGET_SVE_F64MM): New macros.
986 * config/aarch64/iterators.md (SVE_MATMULF): New mode iterator.
987 (UNSPEC_FMMLA, UNSPEC_SMATMUL, UNSPEC_UMATMUL, UNSPEC_USMATMUL)
988 (UNSPEC_TRN1Q, UNSPEC_TRN2Q, UNSPEC_UZP1Q, UNSPEC_UZP2Q, UNSPEC_ZIP1Q)
989 (UNSPEC_ZIP2Q): New unspeccs.
990 (DOTPROD_US_ONLY, PERMUTEQ, MATMUL, FMMLA): New int iterators.
991 (optab, sur, perm_insn): Handle the new unspecs.
992 (sve_fp_op): Handle UNSPEC_FMMLA. Resort.
993 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use
994 TARGET_SVE_F64MM instead of separate tests.
995 (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod<vsi2qi>): New pattern.
996 (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod_lane<vsi2qi>): Likewise.
997 (@aarch64_sve_add_<MATMUL:optab><vsi2qi>): Likewise.
998 (@aarch64_sve_<FMMLA:sve_fp_op><mode>): Likewise.
999 (@aarch64_sve_<PERMUTEQ:optab><mode>): Likewise.
1000 * config/aarch64/aarch64-sve-builtins.cc (TYPES_s_float): New macro.
1001 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): Use it.
1002 (TYPES_s_signed): New macro.
1003 (TYPES_s_integer): Use it.
1004 (TYPES_d_float): New macro.
1005 (TYPES_d_data): Use it.
1006 * config/aarch64/aarch64-sve-builtins-shapes.h (mmla): Declare.
1007 (ternary_intq_uintq_lane, ternary_intq_uintq_opt_n, ternary_uintq_intq)
1008 (ternary_uintq_intq_lane, ternary_uintq_intq_opt_n): Likewise.
1009 * config/aarch64/aarch64-sve-builtins-shapes.cc (mmla_def): New class.
1010 (svmmla): New shape.
1011 (ternary_resize2_opt_n_base): Add TYPE_CLASS2 and TYPE_CLASS3
1012 template parameters.
1013 (ternary_resize2_lane_base): Likewise.
1014 (ternary_resize2_base): New class.
1015 (ternary_qq_lane_base): Likewise.
1016 (ternary_intq_uintq_lane_def): Likewise.
1017 (ternary_intq_uintq_lane): New shape.
1018 (ternary_intq_uintq_opt_n_def): New class
1019 (ternary_intq_uintq_opt_n): New shape.
1020 (ternary_qq_lane_def): Inherit from ternary_qq_lane_base.
1021 (ternary_uintq_intq_def): New class.
1022 (ternary_uintq_intq): New shape.
1023 (ternary_uintq_intq_lane_def): New class.
1024 (ternary_uintq_intq_lane): New shape.
1025 (ternary_uintq_intq_opt_n_def): New class.
1026 (ternary_uintq_intq_opt_n): New shape.
1027 * config/aarch64/aarch64-sve-builtins-base.h (svmmla, svsudot)
1028 (svsudot_lane, svtrn1q, svtrn2q, svusdot, svusdot_lane, svusmmla)
1029 (svuzp1q, svuzp2q, svzip1q, svzip2q): Declare.
1030 * config/aarch64/aarch64-sve-builtins-base.cc (svdot_lane_impl):
1032 (svdotprod_lane_impl): ...this new class.
1033 (svmmla_impl, svusdot_impl): New classes.
1034 (svdot_lane): Update to use svdotprod_lane_impl.
1035 (svmmla, svsudot, svsudot_lane, svtrn1q, svtrn2q, svusdot)
1036 (svusdot_lane, svusmmla, svuzp1q, svuzp2q, svzip1q, svzip2q): New
1038 * config/aarch64/aarch64-sve-builtins-base.def (svmmla): New base
1039 function, with no types defined.
1040 (svmmla, svusmmla, svsudot, svsudot_lane, svusdot, svusdot_lane): New
1041 AARCH64_FL_I8MM functions.
1042 (svmmla): New AARCH64_FL_F32MM function.
1043 (svld1ro): Depend only on AARCH64_FL_F64MM, not on AARCH64_FL_V8_6.
1044 (svmmla, svtrn1q, svtrn2q, svuz1q, svuz2q, svzip1q, svzip2q): New
1045 AARCH64_FL_F64MM function.
1046 (REQUIRED_EXTENSIONS):
1048 2020-01-31 Andrew Stubbs <ams@codesourcery.com>
1050 * config/gcn/gcn-valu.md (addv64di3_exec): Allow one '0' in each
1053 2020-01-31 Uroš Bizjak <ubizjak@gmail.com>
1055 * config/i386/i386.md (*movoi_internal_avx): Do not check for
1056 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL. Remove MODE_V8SF handling.
1057 (*movti_internal): Do not check for
1058 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
1059 (*movtf_internal): Move check for TARGET_SSE2 and size optimization
1060 just after check for TARGET_AVX.
1061 (*movdf_internal): Ditto.
1062 * config/i386/mmx.md (*mov<mode>_internal): Do not check for
1063 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
1064 * config/i386/sse.md (mov<mode>_internal): Only check
1065 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL with V2DFmode. Move check
1066 for TARGET_SSE2 and size optimization just after check for TARGET_AVX.
1067 (<sse>_andnot<mode>3<mask_name>): Move check for
1068 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL after check for TARGET_AVX.
1069 (<code><mode>3<mask_name>): Ditto.
1070 (*andnot<mode>3): Ditto.
1071 (*andnottf3): Ditto.
1072 (*<code><mode>3): Ditto.
1073 (*<code>tf3): Ditto.
1074 (*andnot<VI:mode>3): Remove
1075 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL handling.
1076 (<mask_codefor><code><VI48_AVX_AVX512F:mode>3<mask_name>): Ditto.
1077 (*<code><VI12_AVX_AVX512F:mode>3): Ditto.
1078 (sse4_1_blendv<ssemodesuffix>): Ditto.
1079 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL):
1080 Explain that tune applies to 128bit instructions only.
1082 2020-01-31 Kwok Cheung Yeung <kcy@codesourcery.com>
1084 * config/gcn/mkoffload.c (process_asm): Add sgpr_count and vgpr_count
1085 to definition of hsa_kernel_description. Parse assembly to find SGPR
1086 and VGPR count of kernel and store in hsa_kernel_description.
1088 2020-01-31 Tamar Christina <tamar.christina@arm.com>
1090 PR rtl-optimization/91838
1091 * simplify-rtx.c (simplify_binary_operation_1): Update LSHIFTRT case
1092 to truncate if allowed or reject combination.
1094 2020-01-31 Andrew Stubbs <ams@codesourcery.com>
1096 * tree-ssa-loop-ivopts.c (get_iv): Use sizetype for zero-step.
1097 (find_inv_vars_cb): Likewise.
1099 2020-01-31 David Malcolm <dmalcolm@redhat.com>
1101 * calls.c (special_function_p): Split out the check for DECL_NAME
1102 being non-NULL and fndecl being extern at file scope into a
1103 new maybe_special_function_p and call it. Drop check for fndecl
1104 being non-NULL that was after a usage of DECL_NAME (fndecl).
1105 * tree.h (maybe_special_function_p): New inline function.
1107 2020-01-30 Andrew Stubbs <ams@codesourcery.com>
1109 * config/gcn/gcn-valu.md (gather<mode>_exec): Move contents ...
1110 (mask_gather_load<mode>): ... here, and zero-initialize the
1112 (maskload<mode>di): Zero-initialize the destination.
1115 2020-01-30 David Malcolm <dmalcolm@redhat.com>
1118 * doc/analyzer.texi (Limitations): Note that constraints on
1119 floating-point values are currently ignored.
1121 2020-01-30 Jakub Jelinek <jakub@redhat.com>
1124 * symtab.c (symtab_node::noninterposable_alias): If localalias
1125 already exists, but is not usable, append numbers after it until
1126 a unique name is found. Formatting fix.
1129 * combine.c (simplify_comparison) <case ROTATE>: Punt on out of range
1132 2020-01-30 Andrew Stubbs <ams@codesourcery.com>
1134 * config/gcn/gcn.c (print_operand): Handle LTGT.
1135 * config/gcn/predicates.md (gcn_fp_compare_operator): Allow ltgt.
1137 2020-01-30 Richard Biener <rguenther@suse.de>
1139 * tree-pretty-print.c (dump_generic_node): Wrap VECTOR_CST
1140 and CONSTRUCTOR in _Literal (type) with TDF_GIMPLE.
1142 2020-01-30 John David Anglin <danglin@gcc.gnu.org>
1144 * config/pa/pa.c (pa_elf_select_rtx_section): Place function pointers
1145 without a DECL in .data.rel.ro.local.
1147 2020-01-30 Jakub Jelinek <jakub@redhat.com>
1150 * config/arm/arm.md (uaddvdi4): Actually emit what gen_uaddvsi4
1154 * config/i386/sse.md
1155 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext): Renamed to ...
1156 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext): ... this. Use
1157 any_extend code iterator instead of always zero_extend.
1158 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_lt): Renamed to ...
1159 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_lt): ... this.
1160 Use any_extend code iterator instead of always zero_extend.
1161 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_shift): Renamed to ...
1162 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_shift): ... this.
1163 Use any_extend code iterator instead of always zero_extend.
1164 (*sse2_pmovmskb_ext): New define_insn.
1165 (*sse2_pmovmskb_ext_lt): New define_insn_and_split.
1168 * config/i386/i386.md (*popcountsi2_zext): New define_insn_and_split.
1169 (*popcountsi2_zext_falsedep): New define_insn.
1171 2020-01-30 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
1173 * config.in: Regenerated.
1174 * configure: Regenerated.
1176 2020-01-29 Tobias Burnus <tobias@codesourcery.com>
1179 * config/gcn/gcn-hsa.h (ASM_SPEC): Add -mattr=-code-object-v3 as
1180 LLVM's assembler changed the default in version 9.
1182 2020-01-24 Jeff Law <law@redhat.com>
1184 PR tree-optimization/89689
1185 * builtins.def (BUILT_IN_OBJECT_SIZE): Make it const rather than pure.
1187 2020-01-29 Richard Sandiford <richard.sandiford@arm.com>
1191 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1193 PR rtl-optimization/87763
1194 * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
1195 simplification to handle subregs as well as bare regs.
1196 * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
1198 2020-01-29 Joel Hutton <Joel.Hutton@arm.com>
1201 * ira.c (ira): Revert use of simplified LRA algorithm.
1203 2020-01-29 Martin Jambor <mjambor@suse.cz>
1205 PR tree-optimization/92706
1206 * tree-sra.c (struct access): Fields first_link, last_link,
1207 next_queued and grp_queued renamed to first_rhs_link, last_rhs_link,
1208 next_rhs_queued and grp_rhs_queued respectively, new fields
1209 first_lhs_link, last_lhs_link, next_lhs_queued and grp_lhs_queued.
1210 (struct assign_link): Field next renamed to next_rhs, new field
1211 next_lhs. Updated comment.
1212 (work_queue_head): Renamed to rhs_work_queue_head.
1213 (lhs_work_queue_head): New variable.
1214 (add_link_to_lhs): New function.
1215 (relink_to_new_repr): Also relink LHS lists.
1216 (add_access_to_work_queue): Renamed to add_access_to_rhs_work_queue.
1217 (add_access_to_lhs_work_queue): New function.
1218 (pop_access_from_work_queue): Renamed to
1219 pop_access_from_rhs_work_queue.
1220 (pop_access_from_lhs_work_queue): New function.
1221 (build_accesses_from_assign): Also add links to LHS lists and to LHS
1223 (child_would_conflict_in_lacc): Renamed to
1224 child_would_conflict_in_acc. Adjusted parameter names.
1225 (create_artificial_child_access): New parameter set_grp_read, use it.
1226 (subtree_mark_written_and_enqueue): Renamed to
1227 subtree_mark_written_and_rhs_enqueue.
1228 (propagate_subaccesses_across_link): Renamed to
1229 propagate_subaccesses_from_rhs.
1230 (propagate_subaccesses_from_lhs): New function.
1231 (propagate_all_subaccesses): Also propagate subaccesses from LHSs to
1234 2020-01-29 Martin Jambor <mjambor@suse.cz>
1236 PR tree-optimization/92706
1237 * tree-sra.c (struct access): Adjust comment of
1238 grp_total_scalarization.
1239 (find_access_in_subtree): Look for single children spanning an entire
1241 (scalarizable_type_p): Allow register accesses, adjust callers.
1242 (completely_scalarize): Remove function.
1243 (scalarize_elem): Likewise.
1244 (create_total_scalarization_access): Likewise.
1245 (sort_and_splice_var_accesses): Do not track total scalarization
1247 (analyze_access_subtree): New parameter totally, adjust to new meaning
1248 of grp_total_scalarization.
1249 (analyze_access_trees): Pass new parameter to analyze_access_subtree.
1250 (can_totally_scalarize_forest_p): New function.
1251 (create_total_scalarization_access): Likewise.
1252 (create_total_access_and_reshape): Likewise.
1253 (total_should_skip_creating_access): Likewise.
1254 (totally_scalarize_subtree): Likewise.
1255 (analyze_all_variable_accesses): Perform total scalarization after
1256 subaccess propagation using the new functions above.
1257 (initialize_constant_pool_replacements): Output initializers by
1258 traversing the access tree.
1260 2020-01-29 Martin Jambor <mjambor@suse.cz>
1262 * tree-sra.c (verify_sra_access_forest): New function.
1263 (verify_all_sra_access_forests): Likewise.
1264 (create_artificial_child_access): Set parent.
1265 (analyze_all_variable_accesses): Call the verifier.
1267 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1269 * cgraph.c (cgraph_edge::resolve_speculation): Only lookup direct edge
1270 if called on indirect edge.
1271 (cgraph_edge::redirect_call_stmt_to_callee): Lookup indirect edge of
1272 speculative call if needed.
1274 2020-01-29 Richard Biener <rguenther@suse.de>
1276 PR tree-optimization/93428
1277 * tree-vect-slp.c (vect_build_slp_tree_2): Compute the load
1278 permutation when the load node is created.
1279 (vect_analyze_slp_instance): Re-use it here.
1281 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1283 * ipa-prop.c (update_indirect_edges_after_inlining): Fix warning.
1285 2020-01-28 Vladimir Makarov <vmakarov@redhat.com>
1287 PR rtl-optimization/93272
1288 * ira-lives.c (process_out_of_region_eh_regs): New function.
1289 (process_bb_node_lives): Call it.
1291 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1293 * coverage.c (read_counts_file): Make error message lowercase.
1295 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1297 * profile-count.c (profile_quality_display_names): Fix ordering.
1299 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1302 * cgraph.c (cgraph_add_edge_to_call_site_hash): Update call site
1303 hash only when edge is first within the sequence.
1304 (cgraph_edge::set_call_stmt): Update handling of speculative calls.
1305 (symbol_table::create_edge): Do not set target_prob.
1306 (cgraph_edge::remove_caller): Watch for speculative calls when updating
1308 (cgraph_edge::make_speculative): Drop target_prob parameter.
1309 (cgraph_edge::speculative_call_info): Remove.
1310 (cgraph_edge::first_speculative_call_target): New member function.
1311 (update_call_stmt_hash_for_removing_direct_edge): New function.
1312 (cgraph_edge::resolve_speculation): Rewrite to new API.
1313 (cgraph_edge::speculative_call_for_target): New member function.
1314 (cgraph_edge::make_direct): Rewrite to new API; fix handling of
1315 multiple speculation targets.
1316 (cgraph_edge::redirect_call_stmt_to_callee): Likewise; fix updating
1318 (verify_speculative_call): Verify that targets form an interval.
1319 * cgraph.h (cgraph_edge::speculative_call_info): Remove.
1320 (cgraph_edge::first_speculative_call_target): New member function.
1321 (cgraph_edge::next_speculative_call_target): New member function.
1322 (cgraph_edge::speculative_call_target_ref): New member function.
1323 (cgraph_edge;:speculative_call_indirect_edge): New member funtion.
1324 (cgraph_edge): Remove target_prob.
1325 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
1326 Fix handling of speculative calls.
1327 * ipa-devirt.c (ipa_devirt): Fix handling of speculative cals.
1328 * ipa-fnsummary.c (analyze_function_body): Likewise.
1329 * ipa-inline.c (speculation_useful_p): Use new speculative call API.
1330 * ipa-profile.c (dump_histogram): Fix formating.
1331 (ipa_profile_generate_summary): Watch for overflows.
1332 (ipa_profile): Do not require probablity to be 1/2; update to new API.
1333 * ipa-prop.c (ipa_make_edge_direct_to_target): Update to new API.
1334 (update_indirect_edges_after_inlining): Update to new API.
1335 * ipa-utils.c (ipa_merge_profiles): Rewrite merging of speculative call
1337 * profile-count.h: (profile_probability::adjusted): New.
1338 * tree-inline.c (copy_bb): Update to new speculative call API; fix
1339 updating of profile.
1340 * value-prof.c (gimple_ic_transform): Rename to ...
1341 (dump_ic_profile): ... this one; update dumping.
1342 (stream_in_histogram_value): Fix formating.
1343 (gimple_value_profile_transformations): Update.
1345 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
1348 * config/i386/i386.md (*movoi_internal_avx): Remove
1349 TARGET_SSE_TYPELESS_STORES check.
1350 (*movti_internal): Prefer TARGET_AVX over
1351 TARGET_SSE_TYPELESS_STORES.
1352 (*movtf_internal): Likewise.
1353 * config/i386/sse.md (mov<mode>_internal): Prefer TARGET_AVX over
1354 TARGET_SSE_TYPELESS_STORES. Remove "<MODE_SIZE> == 16" check
1355 from TARGET_SSE_TYPELESS_STORES.
1357 2020-01-28 David Malcolm <dmalcolm@redhat.com>
1359 * diagnostic-core.h (warning_at): Rename overload to...
1360 (warning_meta): ...this.
1361 (emit_diagnostic_valist): Delete decl of overload taking
1362 diagnostic_metadata.
1363 * diagnostic.c (emit_diagnostic_valist): Likewise for defn.
1364 (warning_at): Rename overload taking diagnostic_metadata to...
1365 (warning_meta): ...this.
1367 2020-01-28 Richard Biener <rguenther@suse.de>
1369 PR tree-optimization/93439
1370 * tree-parloops.c (create_loop_fn): Move clique bookkeeping...
1371 * tree-cfg.c (move_sese_region_to_fn): ... here.
1372 (verify_types_in_gimple_reference): Verify used cliques are
1375 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
1378 * config/i386/i386-options.c (set_ix86_tune_features): Add an
1379 argument of a pointer to struct gcc_options and pass it to
1380 parse_mtune_ctrl_str.
1381 (ix86_function_specific_restore): Pass opts to
1382 set_ix86_tune_features.
1383 (ix86_option_override_internal): Likewise.
1384 (parse_mtune_ctrl_str): Add an argument of a pointer to struct
1385 gcc_options and use it for x_ix86_tune_ctrl_string.
1387 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1389 PR rtl-optimization/87763
1390 * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
1391 simplification to handle subregs as well as bare regs.
1392 * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
1394 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1396 * tree-vect-loop.c (vectorizable_reduction): Fail gracefully
1397 for reduction chains that (now) include a call.
1399 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1401 PR tree-optimization/92822
1402 * tree-ssa-forwprop.c (simplify_vector_constructor): When filling
1403 out the don't-care elements of a vector whose significant elements
1404 are duplicates, make the don't-care elements duplicates too.
1406 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1408 PR tree-optimization/93434
1409 * tree-predcom.c (split_data_refs_to_components): Record which
1410 components have had aliasing loads removed. Prevent store-store
1411 commoning for all such components.
1413 2020-01-28 Jakub Jelinek <jakub@redhat.com>
1416 * config/i386/i386.c (ix86_fold_builtin) <do_shift>: If mask is not
1417 -1 or is_vshift is true, use new_vector with number of elts npatterns
1418 rather than new_unary_operation.
1420 PR tree-optimization/93454
1421 * gimple-fold.c (fold_array_ctor_reference): Perform
1422 elt_size.to_uhwi () just once, instead of calling it in every
1423 iteration. Punt if that value is above size of the temporary
1424 buffer. Decrease third native_encode_expr argument when
1425 bufoff + elt_sz is above size of buf.
1427 2020-01-27 Joseph Myers <joseph@codesourcery.com>
1429 * config/mips/mips.c (mips_declare_object_name)
1430 [USE_GNU_UNIQUE_OBJECT]: Support use of gnu_unique_object.
1432 2020-01-27 Martin Liska <mliska@suse.cz>
1434 PR gcov-profile/93403
1435 * tree-profile.c (gimple_init_gcov_profiler): Generate
1436 both __gcov_indirect_call_profiler_v4 and
1437 __gcov_indirect_call_profiler_v4_atomic.
1439 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1442 * config/aarch64/aarch64-simd.md (aarch64_get_half<mode>): New
1444 (@aarch64_split_simd_mov<mode>): Use it.
1445 (aarch64_simd_mov_from_<mode>low): Add a GPR alternative.
1446 Leave the vec_extract patterns to handle 2-element vectors.
1447 (aarch64_simd_mov_from_<mode>high): Likewise.
1448 (vec_extract<VQMOV_NO2E:mode><Vhalf>): New expander.
1449 (vec_extractv2dfv1df): Likewise.
1451 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1453 * config/aarch64/aarch64.c (aarch64_if_then_else_costs): Match
1454 jump conditions for *compare_condjump<GPI:mode>.
1456 2020-01-27 David Malcolm <dmalcolm@redhat.com>
1459 * digraph.cc (test_edge::test_edge): Specify template for base
1462 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
1464 * config/arc/arc.c (arc_rtx_costs): Update mul64 cost.
1466 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
1468 * config/arc/arc-protos.h (gen_mlo): Remove.
1469 (gen_mhi): Likewise.
1470 * config/arc/arc.c (AUX_MULHI): Define.
1471 (arc_must_save_reister): Special handling for r58/59.
1472 (arc_compute_frame_size): Consider mlo/mhi registers.
1473 (arc_save_callee_saves): Emit fp/sp move only when emit_move
1475 (arc_conditional_register_usage): Remove TARGET_BIG_ENDIAN from
1476 mlo/mhi name selection.
1477 (arc_restore_callee_saves): Don't early restore blink when ISR.
1478 (arc_expand_prologue): Add mlo/mhi saving.
1479 (arc_expand_epilogue): Add mlo/mhi restoring.
1482 * config/arc/arc.h (DBX_REGISTER_NUMBER): Correct register
1483 numbering when MUL64 option is used.
1484 (DWARF2_FRAME_REG_OUT): Define.
1485 * config/arc/arc.md (arc600_stall): New pattern.
1486 (VUNSPEC_ARC_ARC600_STALL): Define.
1487 (mulsi64): Use correct mlo/mhi registers.
1488 (mulsi_600): Clean it up.
1489 * config/arc/predicates.md (mlo_operand): Remove any dependency on
1491 (mhi_operand): Likewise.
1493 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
1494 Petro Karashchenko <petro.karashchenko@ring.com>
1496 * config/arc/arc.c (arc_is_uncached_mem_p): Check struct
1497 attributes if needed.
1498 (prepare_move_operands): Generate special unspec instruction for
1500 (arc_isuncached_mem_p): Propagate uncached attribute to each
1502 * config/arc/arc.md (VUNSPEC_ARC_LDDI): Define.
1503 (VUNSPEC_ARC_STDI): Likewise.
1504 (ALLI): New mode iterator.
1505 (mALLI): New mode attribute.
1506 (lddi): New instruction pattern.
1508 (stdidi_split): Split instruction for architectures which are not
1509 supporting ll64 option.
1510 (lddidi_split): Likewise.
1512 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1514 PR rtl-optimization/92989
1515 * lra-lives.c (process_bb_lives): Update the live-in set before
1516 processing additional clobbers.
1518 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1520 PR rtl-optimization/93170
1521 * cselib.c (cselib_invalidate_regno_val): New function, split out
1523 (cselib_invalidate_regno): ...here.
1524 (cselib_invalidated_by_call_p): New function.
1525 (cselib_process_insn): Iterate over all the hard-register entries in
1526 REG_VALUES and invalidate any that cross call-clobbered registers.
1528 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1530 * dojump.c (split_comparison): Use HONOR_NANS rather than
1531 HONOR_SNANS when splitting LTGT.
1533 2020-01-27 Martin Liska <mliska@suse.cz>
1536 * opts.c (print_filtered_help): Exclude language-specific
1537 options from --help=common unless enabled in all FEs.
1539 2020-01-27 Martin Liska <mliska@suse.cz>
1541 * opts.c (print_help): Exclude params from
1542 all except --help=param.
1544 2020-01-27 Martin Liska <mliska@suse.cz>
1547 * config/i386/i386-features.c (make_resolver_func):
1548 Align the code with ppc64 target implementation.
1549 Do not generate a unique name for resolver function.
1551 2020-01-27 Richard Biener <rguenther@suse.de>
1553 PR tree-optimization/93397
1554 * tree-vect-slp.c (vect_analyze_slp_instance): Delay
1555 converted reduction chain SLP graph adjustment.
1557 2020-01-26 Marek Polacek <polacek@redhat.com>
1560 * sanopt.c (sanitize_rewrite_addressable_params): Avoid crash on
1563 2020-01-26 Jason Merrill <jason@redhat.com>
1566 * tree.c (verify_type_variant): Only verify TYPE_NEEDS_CONSTRUCTING
1569 2020-01-26 Darius Galis <darius.galis@cyberthorstudios.com>
1571 * config/rx/rx.md (setmemsi): Added rx_allow_string_insns constraint
1572 (rx_setmem): Likewise.
1574 2020-01-26 Jakub Jelinek <jakub@redhat.com>
1577 * config/i386/i386.md (*addv<dwi>4_doubleword, *subv<dwi>4_doubleword):
1578 Use nonimmediate_operand instead of x86_64_hilo_general_operand and
1579 drop <di> from constraint of last operand.
1582 * config/i386/sse.md (*avx_vperm_broadcast_<mode>): Disallow for
1583 TARGET_AVX2 and V4DFmode not in the split condition, but in the
1584 pattern condition, though allow { 0, 0, 0, 0 } broadcast always.
1586 2020-01-25 Feng Xue <fxue@os.amperecomputing.com>
1589 * ipa-cp.c (get_info_about_necessary_edges): Remove value
1592 2020-01-24 Jeff Law <law@redhat.com>
1594 PR tree-optimization/92788
1595 * tree-ssa-threadedge.c (thread_across_edge): Check EDGE_COMPLEX
1598 2020-01-24 Jakub Jelinek <jakub@redhat.com>
1601 * config/i386/sse.md (*avx_vperm_broadcast_v4sf,
1602 *avx_vperm_broadcast_<mode>,
1603 <sse2_avx_avx512f>_vpermil<mode><mask_name>,
1604 *<sse2_avx_avx512f>_vpermilp<mode><mask_name>):
1605 Move before avx2_perm<mode>/avx512f_perm<mode>.
1608 * simplify-rtx.c (simplify_const_unary_operation,
1609 simplify_const_binary_operation): Punt for mode precision above
1610 MAX_BITSIZE_MODE_ANY_INT.
1612 2020-01-24 Andrew Pinski <apinski@marvell.com>
1614 * config/arm/aarch-cost-tables.h (cortexa57_extra_costs): Change
1617 2020-01-24 Jeff Law <law@redhat.com>
1620 * config/h8300/h8300.c (h8300_print_operand): Only call byte_reg
1621 for REGs. Call output_operand_lossage to get more reasonable
1624 2020-01-24 Andrew Stubbs <ams@codesourcery.com>
1626 * config/gcn/gcn-valu.md (vec_cmp<mode>di): Use
1627 gcn_fp_compare_operator.
1628 (vec_cmpu<mode>di): Use gcn_compare_operator.
1629 (vec_cmp<u>v64qidi): Use gcn_compare_operator.
1630 (vec_cmp<mode>di_exec): Use gcn_fp_compare_operator.
1631 (vec_cmpu<mode>di_exec): Use gcn_compare_operator.
1632 (vec_cmp<u>v64qidi_exec): Use gcn_compare_operator.
1633 (vec_cmp<mode>di_dup): Use gcn_fp_compare_operator.
1634 (vec_cmp<mode>di_dup_exec): Use gcn_fp_compare_operator.
1635 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): Use
1636 gcn_fp_compare_operator.
1637 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): Use
1638 gcn_fp_compare_operator.
1639 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): Use
1640 gcn_fp_compare_operator.
1641 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): Use
1642 gcn_fp_compare_operator.
1644 2020-01-24 Maciej W. Rozycki <macro@wdc.com>
1646 * doc/install.texi (Cross-Compiler-Specific Options): Document
1647 `--with-toolexeclibdir' option.
1649 2020-01-24 Hans-Peter Nilsson <hp@axis.com>
1651 * target.def (flags_regnum): Also mention effect on delay slot filling.
1652 * doc/tm.texi: Regenerate.
1654 2020-01-23 Jeff Law <law@redhat.com>
1656 PR translation/90162
1657 * config/h8300/h8300.c (h8300_option_override): Fix diagnostic text.
1659 2020-01-23 Mikael Tillenius <mti-1@tillenius.com>
1662 * config/h8300/h8300.h (FUNCTION_PROFILER): Fix emission of
1665 2020-01-23 Jakub Jelinek <jakub@redhat.com>
1667 PR rtl-optimization/93402
1668 * postreload.c (reload_combine_recognize_pattern): Don't try to adjust
1671 2020-01-23 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
1673 * config.in: Regenerated.
1674 * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to 1
1675 for TARGET_LIBC_GNUSTACK.
1676 * configure: Regenerated.
1677 * configure.ac: Define TARGET_LIBC_GNUSTACK if glibc version is
1678 found to be 2.31 or greater.
1680 2020-01-23 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
1682 * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to
1684 * config/mips/mips.c (TARGET_ASM_FILE_END): Define to ...
1685 (mips_asm_file_end): New function. Delegate to
1686 file_end_indicate_exec_stack if NEED_INDICATE_EXEC_STACK is true.
1687 * config/mips/mips.h (NEED_INDICATE_EXEC_STACK): Define to 0.
1689 2020-01-23 Jakub Jelinek <jakub@redhat.com>
1692 * config/i386/i386-modes.def (POImode): New mode.
1693 (MAX_BITSIZE_MODE_ANY_INT): Change from 128 to 160.
1694 * config/i386/i386.md (DPWI): New mode attribute.
1695 (addv<mode>4, subv<mode>4): Use <DPWI> instead of <DWI>.
1697 (QPWI): ... this. Use POI instead of OI for TImode.
1698 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1,
1699 *subv<dwi>4_doubleword, *subv<dwi>4_doubleword_1): Use <QPWI>
1702 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
1705 * config/aarch64/aarch64.md (UNSPEC_SPECULATION_TRACKER_REV): New
1707 (speculation_tracker_rev): New pattern.
1708 * config/aarch64/aarch64-speculation.cc (aarch64_do_track_speculation):
1709 Use speculation_tracker_rev to track the inverse condition.
1711 2020-01-23 Richard Biener <rguenther@suse.de>
1713 PR tree-optimization/93381
1714 * tree-ssa-sccvn.c (vn_walk_cb_data::push_partial_def): Take
1715 alias-set of the def as argument and record the first one.
1716 (vn_walk_cb_data::first_set): New member.
1717 (vn_reference_lookup_3): Pass the alias-set of the current def
1718 to push_partial_def. Fix alias-set used in the aggregate copy
1720 (vn_reference_lookup): Consistently set *last_vuse_ptr.
1721 * real.c (clear_significand_below): Fix out-of-bound access.
1723 2020-01-23 Jakub Jelinek <jakub@redhat.com>
1726 * config/i386/i386.md (*bmi2_bzhi_<mode>3_2, *bmi2_bzhi_<mode>3_3):
1727 New define_insn patterns.
1729 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
1731 * doc/sourcebuild.texi (check-function-bodies): Add an
1732 optional target/xfail selector.
1734 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
1736 PR rtl-optimization/93124
1737 * auto-inc-dec.c (merge_in_block): Don't add auto inc/decs to
1738 bare USE and CLOBBER insns.
1740 2020-01-22 Andrew Pinski <apinski@marvell.com>
1742 * config/arc/arc.c (output_short_suffix): Check insn for nullness.
1744 2020-01-22 David Malcolm <dmalcolm@redhat.com>
1747 * gdbinit.in (break-on-saved-diagnostic): Update for move of
1748 diagnostic_manager into "ana" namespace.
1749 * selftest-run-tests.c (selftest::run_tests): Update for move of
1750 selftest::run_analyzer_selftests to
1751 ana::selftest::run_analyzer_selftests.
1753 2020-01-22 Richard Sandiford <richard.sandiford@arm.com>
1755 * cfgexpand.c (union_stack_vars): Update the size.
1757 2020-01-22 Richard Biener <rguenther@suse.de>
1759 PR tree-optimization/93381
1760 * tree-ssa-structalias.c (find_func_aliases): Assume offsetting
1761 throughout, handle all conversions the same.
1763 2020-01-22 Jakub Jelinek <jakub@redhat.com>
1766 * config/aarch64/aarch64.c (aarch64_expand_subvti): Only use
1767 gen_subdi3_compare1_imm if low_in2 satisfies aarch64_plus_immediate
1768 predicate, not whenever it is CONST_INT. Otherwise, force_reg it.
1769 Call force_reg on high_in2 unconditionally.
1771 2020-01-22 Martin Liska <mliska@suse.cz>
1773 PR tree-optimization/92924
1774 * profile.c (compute_value_histograms): Divide
1777 2020-01-22 Jakub Jelinek <jakub@redhat.com>
1780 * output.h (assemble_name_resolve): Declare.
1781 * varasm.c (assemble_name_resolve): New function.
1782 (assemble_name): Use it.
1783 * config/i386/i386.h (ASM_OUTPUT_SYMBOL_REF): Define.
1785 2020-01-22 Joseph Myers <joseph@codesourcery.com>
1787 * doc/sourcebuild.texi (Texinfo Manuals, Front End): Refer to
1788 update_web_docs_git instead of update_web_docs_svn.
1790 2020-01-21 Andrew Pinski <apinski@marvell.com>
1793 * config/aarch64/aarch64.md (tlsgd_small_<mode>): Have operand 0
1794 as PTR mode. Have operand 1 as being modeless, it can be P mode.
1795 (*tlsgd_small_<mode>): Likewise.
1796 * config/aarch64/aarch64.c (aarch64_load_symref_appropriately)
1797 <case SYMBOL_SMALL_TLSGD>: Call gen_tlsgd_small_* with a ptr_mode
1798 register. Convert that register back to dest using convert_mode.
1800 2020-01-21 Jim Wilson <jimw@sifive.com>
1802 * config/riscv/riscv-sr.c (riscv_sr_match_prologue): Use INTVAL
1805 2020-01-21 H.J. Lu <hongjiu.lu@intel.com>
1806 Uros Bizjak <ubizjak@gmail.com>
1809 * config/i386/i386.c (ix86_tls_module_base): Replace Pmode
1811 (legitimize_tls_address): Do GNU2 TLS address computation in
1812 ptr_mode and zero-extend result to Pmode.
1813 * config/i386/i386.md (@tls_dynamic_gnu2_64_<mode>): Replace
1814 :P with :PTR and Pmode with ptr_mode.
1815 (*tls_dynamic_gnu2_lea_64_<mode>): Likewise.
1816 (*tls_dynamic_gnu2_call_64_<mode>): Likewise.
1817 (*tls_dynamic_gnu2_combine_64_<mode>): Likewise.
1819 2020-01-21 Jakub Jelinek <jakub@redhat.com>
1822 * config/riscv/riscv.c (riscv_rtx_costs) <case ZERO_EXTRACT>: Verify
1823 the last two operands are CONST_INT_P before using them as such.
1825 2020-01-21 Richard Sandiford <richard.sandiford@arm.com>
1827 * config/aarch64/aarch64-sve-builtins.def: Use get_typenode_from_name
1828 to get the integer element types.
1830 2020-01-21 Richard Sandiford <richard.sandiford@arm.com>
1832 * config/aarch64/aarch64-sve-builtins.h
1833 (function_expander::convert_to_pmode): Declare.
1834 * config/aarch64/aarch64-sve-builtins.cc
1835 (function_expander::convert_to_pmode): New function.
1836 (function_expander::get_contiguous_base): Use it.
1837 (function_expander::prepare_gather_address_operands): Likewise.
1838 * config/aarch64/aarch64-sve-builtins-sve2.cc
1839 (svwhilerw_svwhilewr_impl::expand): Likewise.
1841 2020-01-21 Szabolcs Nagy <szabolcs.nagy@arm.com>
1844 * config/aarch64/aarch64.c (aarch64_declare_function_name): Set
1845 cfun->machine->label_is_assembled.
1846 (aarch64_print_patchable_function_entry): New.
1847 (TARGET_ASM_PRINT_PATCHABLE_FUNCTION_ENTRY): Define.
1848 * config/aarch64/aarch64.h (struct machine_function): New field,
1851 2020-01-21 David Malcolm <dmalcolm@redhat.com>
1854 * ipa-profile.c (ipa_profile): Delete call_sums and set it to
1857 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
1860 * cgraph.c (cgraph_edge::resolve_speculation,
1861 cgraph_edge::redirect_call_stmt_to_callee): Fix update of
1862 call_stmt_site_hash.
1864 2020-01-21 Martin Liska <mliska@suse.cz>
1866 * config/rs6000/rs6000.c (common_mode_defined): Remove
1869 2020-01-21 Richard Biener <rguenther@suse.de>
1871 PR tree-optimization/92328
1872 * tree-ssa-sccvn.c (vn_reference_lookup_3): Preserve
1873 type when value-numbering same-sized store by inserting a
1875 (eliminate_dom_walker::eliminate_stmt): When eliminating
1876 a redundant store handle bit-reinterpretation of the same value.
1878 2020-01-21 Andrew Pinski <apinski@marvel.com>
1881 * tree-into-ssa.c (prepare_block_for_update_1): Split out
1883 (prepare_block_for_update): This. Use a worklist instead of
1886 2020-01-21 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1888 * gcc/config/arm/arm.c (clear_operation_p):
1889 Initialise last_regno, skip first iteration
1890 based on the first_set value and use ints instead
1891 of the unnecessary HOST_WIDE_INTs.
1893 2020-01-21 Jakub Jelinek <jakub@redhat.com>
1896 * config/rs6000/rs6000.c (rs6000_emit_cmove): If using fsel, punt for
1897 compare_mode other than SFmode or DFmode.
1899 2020-01-21 Kito Cheng <kito.cheng@sifive.com>
1902 * config/riscv/riscv-protos.h (riscv_hard_regno_rename_ok): New.
1903 * config/riscv/riscv.c (riscv_hard_regno_rename_ok): New.
1904 * config/riscv/riscv.h (HARD_REGNO_RENAME_OK): Defined.
1906 2020-01-20 Wilco Dijkstra <wdijkstr@arm.com>
1908 * config/aarch64/aarch64.c (neoversen1_tunings): Set jump_align to 4.
1910 2020-01-20 Andrew Pinski <apinski@marvell.com>
1913 * targhooks.c (default_print_patchable_function_entry): Use
1914 output_asm_insn to emit the nop instruction.
1916 2020-01-20 Fangrui Song <maskray@google.com>
1919 * targhooks.c (default_print_patchable_function_entry): Align to
1922 2020-01-20 H.J. Lu <hongjiu.lu@intel.com>
1925 * config/i386/i386.c (legitimize_tls_address): Pass Pmode to
1926 gen_tls_dynamic_gnu2_64. Compute GNU2 TLS address in ptr_mode.
1927 * config/i386/i386.md (tls_dynamic_gnu2_64): Renamed to ...
1928 (@tls_dynamic_gnu2_64_<mode>): This. Replace DI with P.
1929 (*tls_dynamic_gnu2_lea_64): Renamed to ...
1930 (*tls_dynamic_gnu2_lea_64_<mode>): This. Replace DI with P.
1931 Remove the {q} suffix from lea.
1932 (*tls_dynamic_gnu2_call_64): Renamed to ...
1933 (*tls_dynamic_gnu2_call_64_<mode>): This. Replace DI with P.
1934 (*tls_dynamic_gnu2_combine_64): Renamed to ...
1935 (*tls_dynamic_gnu2_combine_64_<mode>): This. Replace DI with P.
1936 Pass Pmode to gen_tls_dynamic_gnu2_64.
1938 2020-01-20 Wilco Dijkstra <wdijkstr@arm.com>
1940 * config/aarch64/aarch64.h (SLOW_BYTE_ACCESS): Set to 1.
1942 2020-01-20 Richard Sandiford <richard.sandiford@arm.com>
1944 * config/aarch64/aarch64-sve-builtins-base.cc
1945 (svld1ro_impl::memory_vector_mode): Remove parameter name.
1947 2020-01-20 Richard Biener <rguenther@suse.de>
1950 * dwarf2out.c (prune_unused_types): Unconditionally mark
1951 called function DIEs.
1953 2020-01-20 Martin Liska <mliska@suse.cz>
1955 PR tree-optimization/93199
1956 * tree-eh.c (struct leh_state): Add
1957 new field outer_non_cleanup.
1958 (cleanup_is_dead_in): Pass leh_state instead
1959 of eh_region. Add a checking that state->outer_non_cleanup
1960 points to outer non-clean up region.
1961 (lower_try_finally): Record outer_non_cleanup
1963 (lower_catch): Likewise.
1964 (lower_eh_filter): Likewise.
1965 (lower_eh_must_not_throw): Likewise.
1966 (lower_cleanup): Likewise.
1968 2020-01-20 Richard Biener <rguenther@suse.de>
1970 PR tree-optimization/93094
1971 * tree-vectorizer.h (vect_loop_versioning): Adjust.
1972 (vect_transform_loop): Likewise.
1973 * tree-vectorizer.c (try_vectorize_loop_1): Pass down
1974 loop_vectorized_call to vect_transform_loop.
1975 * tree-vect-loop.c (vect_transform_loop): Pass down
1976 loop_vectorized_call to vect_loop_versioning.
1977 * tree-vect-loop-manip.c (vect_loop_versioning): Use
1978 the earlier discovered loop_vectorized_call.
1980 2020-01-19 Eric S. Raymond <esr@thyrsus.com>
1982 * doc/contribute.texi: Update for SVN -> Git transition.
1983 * doc/install.texi: Likewise.
1985 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
1988 * cgraph.c (cgraph_edge::make_speculative): Increase number of
1989 speculative targets.
1990 (verify_speculative_call): New function
1991 (cgraph_node::verify_node): Use it.
1992 * ipa-profile.c (ipa_profile): Fix formating; do not set number of
1995 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
1998 * cgraph.c (cgraph_edge::resolve_speculation): Fix foramting.
1999 (cgraph_edge::make_direct): Remove all indirect targets.
2000 (cgraph_edge::redirect_call_stmt_to_callee): Use make_direct..
2001 (cgraph_node::verify_node): Verify that only one call_stmt or
2002 lto_stmt_uid is set.
2003 * cgraphclones.c (cgraph_edge::clone): Set only one call_stmt or
2005 * lto-cgraph.c (lto_output_edge): Simplify streaming of stmt.
2006 (lto_output_ref): Simplify streaming of stmt.
2007 * lto-streamer-in.c (fixup_call_stmt_edges_1): Clear lto_stmt_uid.
2009 2020-01-18 Tamar Christina <tamar.christina@arm.com>
2011 * config/aarch64/aarch64-sve-builtins-base.cc (memory_vector_mode):
2012 Mark parameter unused.
2014 2020-01-18 Hans-Peter Nilsson <hp@axis.com>
2016 * config.gcc <obsolete targets>: Add crisv32-*-* and cris-*-linux*
2018 2019-01-18 Gerald Pfeifer <gerald@pfeifer.com>
2020 * varpool.c (ctor_useable_for_folding_p): Fix grammar.
2022 2020-01-18 Iain Sandoe <iain@sandoe.co.uk>
2024 * Makefile.in: Add coroutine-passes.o.
2025 * builtin-types.def (BT_CONST_SIZE): New.
2026 (BT_FN_BOOL_PTR): New.
2027 (BT_FN_PTR_PTR_CONST_SIZE_BOOL): New.
2028 * builtins.def (DEF_COROUTINE_BUILTIN): New.
2029 * coroutine-builtins.def: New file.
2030 * coroutine-passes.cc: New file.
2031 * function.h (struct GTY function): Add a bit to indicate that the
2032 function is a coroutine component.
2033 * internal-fn.c (expand_CO_FRAME): New.
2034 (expand_CO_YIELD): New.
2035 (expand_CO_SUSPN): New.
2036 (expand_CO_ACTOR): New.
2037 * internal-fn.def (CO_ACTOR): New.
2041 * passes.def: Add pass_coroutine_lower_builtins,
2042 pass_coroutine_early_expand_ifns.
2043 * tree-pass.h (make_pass_coroutine_lower_builtins): New.
2044 (make_pass_coroutine_early_expand_ifns): New.
2045 * doc/invoke.texi: Document the fcoroutines command line
2048 2020-01-18 Jakub Jelinek <jakub@redhat.com>
2050 * config/arm/vfp.md (*clear_vfp_multiple): Remove unused variable.
2053 * config/arm/arm.c (clear_operation_p): Don't use REGNO until
2054 after checking the argument is a REG. Don't use REGNO (reg)
2055 again to set last_regno, reuse regno variable instead.
2057 2020-01-17 David Malcolm <dmalcolm@redhat.com>
2059 * doc/analyzer.texi (Limitations): Add note about NaN.
2061 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2062 Sudakshina Das <sudi.das@arm.com>
2064 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for both reg
2065 and valid immediate.
2066 (ashrdi3): Generate thumb2_asrl for both reg and valid immediate.
2067 (lshrdi3): Generate thumb2_lsrl for valid immediates.
2068 * config/arm/constraints.md (Pg): New.
2069 * config/arm/predicates.md (long_shift_imm): New.
2070 (arm_reg_or_long_shift_imm): Likewise.
2071 * config/arm/thumb2.md (thumb2_asrl): New immediate alternative.
2072 (thumb2_lsll): Likewise.
2075 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2076 Sudakshina Das <sudi.das@arm.com>
2078 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for TARGET_HAVE_MVE.
2079 (ashrdi3): Generate thumb2_asrl for TARGET_HAVE_MVE.
2080 * config/arm/arm.c (arm_hard_regno_mode_ok): Allocate even odd
2081 register pairs for doubleword quantities for ARMv8.1M-Mainline.
2082 * config/arm/thumb2.md (thumb2_asrl): New.
2083 (thumb2_lsll): Likewise.
2085 2020-01-17 Jakub Jelinek <jakub@redhat.com>
2087 * config/arm/arm.c (cmse_nonsecure_call_inline_register_clear): Remove
2090 2020-01-17 Alexander Monakov <amonakov@ispras.ru>
2092 * gdbinit.in (help-gcc-hooks): New command.
2093 (pp, pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, ptc, pdn, ptn, pdd, prc,
2094 pi, pbm, pel, trt): Take $arg0 instead of $ if supplied. Update
2097 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
2099 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use the
2100 correct target macro.
2102 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
2104 * config/aarch64/aarch64-protos.h
2105 (aarch64_sve_ld1ro_operand_p): New.
2106 * config/aarch64/aarch64-sve-builtins-base.cc
2107 (class load_replicate): New.
2108 (class svld1ro_impl): New.
2109 (class svld1rq_impl): Change to inherit from load_replicate.
2110 (svld1ro): New sve intrinsic function base.
2111 * config/aarch64/aarch64-sve-builtins-base.def (svld1ro):
2112 New DEF_SVE_FUNCTION.
2113 * config/aarch64/aarch64-sve-builtins-base.h
2114 (svld1ro): New decl.
2115 * config/aarch64/aarch64-sve-builtins.cc
2116 (function_expander::add_mem_operand): Modify assert to allow
2118 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): New
2120 * config/aarch64/aarch64.c
2121 (aarch64_sve_ld1rq_operand_p): Implement in terms of ...
2122 (aarch64_sve_ld1rq_ld1ro_operand_p): This.
2123 (aarch64_sve_ld1ro_operand_p): New.
2124 * config/aarch64/aarch64.md (UNSPEC_LD1RO): New unspec.
2125 * config/aarch64/constraints.md (UOb,UOh,UOw,UOd): New.
2126 * config/aarch64/predicates.md
2127 (aarch64_sve_ld1ro_operand_{b,h,w,d}): New.
2129 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
2131 * config/aarch64/aarch64-c.c (_ARM_FEATURE_MATMUL_FLOAT64):
2132 Introduce this ACLE specified predefined macro.
2133 * config/aarch64/aarch64-option-extensions.def (f64mm): New.
2134 (fp): Disabling this disables f64mm.
2135 (simd): Disabling this disables f64mm.
2136 (fp16): Disabling this disables f64mm.
2137 (sve): Disabling this disables f64mm.
2138 * config/aarch64/aarch64.h (AARCH64_FL_F64MM): New.
2139 (AARCH64_ISA_F64MM): New.
2140 (TARGET_F64MM): New.
2141 * doc/invoke.texi (f64mm): Document new option.
2143 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
2145 * config/aarch64/aarch64.c (generic_tunings): Add branch fusion.
2146 (neoversen1_tunings): Likewise.
2148 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
2151 * config/aarch64/aarch64.c (aarch64_split_compare_and_swap)
2152 Add assert to ensure prolog has been emitted.
2153 (aarch64_split_atomic_op): Likewise.
2154 * config/aarch64/atomics.md (aarch64_compare_and_swap<mode>)
2155 Use epilogue_completed rather than reload_completed.
2156 (aarch64_atomic_exchange<mode>): Likewise.
2157 (aarch64_atomic_<atomic_optab><mode>): Likewise.
2158 (atomic_nand<mode>): Likewise.
2159 (aarch64_atomic_fetch_<atomic_optab><mode>): Likewise.
2160 (atomic_fetch_nand<mode>): Likewise.
2161 (aarch64_atomic_<atomic_optab>_fetch<mode>): Likewise.
2162 (atomic_nand_fetch<mode>): Likewise.
2164 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
2167 * config/aarch64/aarch64.h (REVERSIBLE_CC_MODE): Return false
2169 (REVERSE_CONDITION): Delete.
2170 * config/aarch64/iterators.md (CC_ONLY): New mode iterator.
2171 (CCFP_CCFPE): Likewise.
2172 (e): New mode attribute.
2173 * config/aarch64/aarch64.md (ccmp<GPI:mode>): Rename to...
2174 (@ccmp<CC_ONLY:mode><GPI:mode>): ...this, using CC_ONLY instead of CC.
2175 (fccmp<GPF:mode>, fccmpe<GPF:mode>): Merge into...
2176 (@ccmp<CCFP_CCFPE:mode><GPF:mode>): ...this combined pattern.
2177 (@ccmp<CC_ONLY:mode><GPI:mode>_rev): New pattern.
2178 (@ccmp<CCFP_CCFPE:mode><GPF:mode>_rev): Likewise.
2179 * config/aarch64/aarch64.c (aarch64_gen_compare_reg): Update
2180 name of generator from gen_ccmpdi to gen_ccmpccdi.
2181 (aarch64_gen_ccmp_next): Use code_for_ccmp. If we want to reverse
2182 the previous comparison but aren't able to, use the new ccmp_rev
2185 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
2187 * gimplify.c (gimplify_return_expr): Use poly_int_tree_p rather
2188 than testing directly for INTEGER_CST.
2189 (gimplify_target_expr, gimplify_omp_depend): Likewise.
2191 2020-01-17 Jakub Jelinek <jakub@redhat.com>
2193 PR tree-optimization/93292
2194 * tree-vect-stmts.c (vectorizable_comparison): Punt also if
2195 get_vectype_for_scalar_type returns NULL.
2197 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
2199 * params.opt (-param=max-predicted-iterations): Increase range from 0.
2200 * predict.c (estimate_loops): Add 1 to param_max_predicted_iterations.
2202 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
2204 * ipa-fnsummary.c (estimate_calls_size_and_time): Fix formating of
2206 * params.opt: (max-predicted-iterations): Set bounds.
2207 * predict.c (real_almost_one, real_br_prob_base,
2208 real_inv_br_prob_base, real_one_half, real_bb_freq_max): Remove.
2209 (propagate_freq): Add max_cyclic_prob parameter; cap cyclic
2210 probabilities; do not truncate to reg_br_prob_bases.
2211 (estimate_loops_at_level): Pass max_cyclic_prob.
2212 (estimate_loops): Compute max_cyclic_prob.
2213 (estimate_bb_frequencies): Do not initialize real_*; update calculation
2215 * profile-count.c (profile_probability::to_sreal): New.
2216 * profile-count.h (class sreal): Move up in file.
2217 (profile_probability::to_sreal): Declare.
2219 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2222 (arm_invalid_conversion): New function for target hook.
2223 (arm_invalid_unary_op): New function for target hook.
2224 (arm_invalid_binary_op): New function for target hook.
2226 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2228 * config.gcc: Add arm_bf16.h.
2229 * config/arm/arm-builtins.c (arm_mangle_builtin_type): Fix comment.
2230 (arm_simd_builtin_std_type): Add BFmode.
2231 (arm_init_simd_builtin_types): Define element types for vector types.
2232 (arm_init_bf16_types): New function.
2233 (arm_init_builtins): Add arm_init_bf16_types function call.
2234 * config/arm/arm-modes.def: Add BFmode and V4BF, V8BF vector modes.
2235 * config/arm/arm-simd-builtin-types.def: Add V4BF, V8BF.
2236 * config/arm/arm.c (aapcs_vfp_sub_candidate): Add BFmode.
2237 (arm_hard_regno_mode_ok): Add BFmode and tidy up statements.
2238 (arm_vector_mode_supported_p): Add V4BF, V8BF.
2239 (arm_mangle_type): Add __bf16.
2240 * config/arm/arm.h: Add V4BF, V8BF to VALID_NEON_DREG_MODE,
2241 VALID_NEON_QREG_MODE respectively. Add export arm_bf16_type_node,
2242 arm_bf16_ptr_type_node.
2243 * config/arm/arm.md: Add BFmode to movhf expand, mov pattern and
2244 define_split between ARM registers.
2245 * config/arm/arm_bf16.h: New file.
2246 * config/arm/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
2247 * config/arm/iterators.md: (ANY64_BF, VDXMOV, VHFBF, HFBF, fporbf): New.
2249 * config/arm/neon.md: Add BF vector types to movhf NEON move patterns.
2250 * config/arm/vfp.md: Add BFmode to movhf patterns.
2252 2020-01-16 Mihail Ionescu <mihail.ionescu@arm.com>
2253 Andre Vieira <andre.simoesdiasvieira@arm.com>
2255 * config/arm/arm-cpus.in (mve, mve_float): New features.
2256 (dsp, mve, mve.fp): New options.
2257 * config/arm/arm.h (TARGET_HAVE_MVE, TARGET_HAVE_MVE_FLOAT): Define.
2258 * config/arm/t-rmprofile: Map v8.1-M multilibs to v8-M.
2259 * doc/invoke.texi: Document the armv8.1-m mve and dps options.
2261 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2262 Thomas Preud'homme <thomas.preudhomme@arm.com>
2264 * config/arm/arm-cpus.in (ARMv8_1m_main): Redefine as an extension to
2266 * config/arm/arm.c (arm_options_perform_arch_sanity_checks): Remove
2267 error for using -mcmse when targeting Armv8.1-M Mainline.
2269 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2270 Thomas Preud'homme <thomas.preudhomme@arm.com>
2272 * config/arm/arm.md (nonsecure_call_internal): Do not force memory
2273 address in r4 when targeting Armv8.1-M Mainline.
2274 (nonsecure_call_value_internal): Likewise.
2275 * config/arm/thumb2.md (nonsecure_call_reg_thumb2): Make memory address
2276 a register match_operand again. Emit BLXNS when targeting
2278 (nonsecure_call_value_reg_thumb2): Likewise.
2280 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2281 Thomas Preud'homme <thomas.preudhomme@arm.com>
2283 * config/arm/arm.c (arm_add_cfa_adjust_cfa_note): Declare early.
2284 (cmse_nonsecure_call_inline_register_clear): Define new lazy_fpclear
2285 variable as true when floating-point ABI is not hard. Replace
2286 check against TARGET_HARD_FLOAT_ABI by checks against lazy_fpclear.
2287 Generate VLSTM and VLLDM instruction respectively before and
2288 after a function call to cmse_nonsecure_call function.
2289 * config/arm/unspecs.md (VUNSPEC_VLSTM): Define unspec.
2290 (VUNSPEC_VLLDM): Likewise.
2291 * config/arm/vfp.md (lazy_store_multiple_insn): New define_insn.
2292 (lazy_load_multiple_insn): Likewise.
2294 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2295 Thomas Preud'homme <thomas.preudhomme@arm.com>
2297 * config/arm/arm.c (vfp_emit_fstmd): Declare early.
2298 (arm_emit_vfp_multi_reg_pop): Likewise.
2299 (cmse_nonsecure_call_inline_register_clear): Abstract number of VFP
2300 registers to clear in max_fp_regno. Emit VPUSH and VPOP to save and
2301 restore callee-saved VFP registers.
2303 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2304 Thomas Preud'homme <thomas.preudhomme@arm.com>
2306 * config/arm/arm.c (arm_emit_multi_reg_pop): Declare early.
2307 (cmse_nonsecure_call_clear_caller_saved): Rename into ...
2308 (cmse_nonsecure_call_inline_register_clear): This. Save and clear
2309 callee-saved GPRs as well as clear ip register before doing a nonsecure
2310 call then restore callee-saved GPRs after it when targeting
2312 (arm_reorg): Adapt to function rename.
2314 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2315 Thomas Preud'homme <thomas.preudhomme@arm.com>
2317 * config/arm/arm-protos.h (clear_operation_p): Adapt prototype.
2318 * config/arm/arm.c (clear_operation_p): Extend to be able to check a
2319 clear_vfp_multiple pattern based on a new vfp parameter.
2320 (cmse_clear_registers): Generate VSCCLRM to clear VFP registers when
2321 targeting Armv8.1-M Mainline.
2322 (cmse_nonsecure_entry_clear_before_return): Clear VFP registers
2323 unconditionally when targeting Armv8.1-M Mainline architecture. Check
2324 whether VFP registers are available before looking call_used_regs for a
2326 * config/arm/predicates.md (clear_multiple_operation): Adapt to change
2327 of prototype of clear_operation_p.
2328 (clear_vfp_multiple_operation): New predicate.
2329 * config/arm/unspecs.md (VUNSPEC_VSCCLRM_VPR): New volatile unspec.
2330 * config/arm/vfp.md (clear_vfp_multiple): New define_insn.
2332 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2333 Thomas Preud'homme <thomas.preudhomme@arm.com>
2335 * config/arm/arm-protos.h (clear_operation_p): Declare.
2336 * config/arm/arm.c (clear_operation_p): New function.
2337 (cmse_clear_registers): Generate clear_multiple instruction pattern if
2338 targeting Armv8.1-M Mainline or successor.
2339 (output_return_instruction): Only output APSR register clearing if
2340 Armv8.1-M Mainline instructions not available.
2341 (thumb_exit): Likewise.
2342 * config/arm/predicates.md (clear_multiple_operation): New predicate.
2343 * config/arm/thumb2.md (clear_apsr): New define_insn.
2344 (clear_multiple): Likewise.
2345 * config/arm/unspecs.md (VUNSPEC_CLRM_APSR): New volatile unspec.
2347 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2348 Thomas Preud'homme <thomas.preudhomme@arm.com>
2350 * config/arm/arm.c (fp_sysreg_names): Declare and define.
2351 (use_return_insn): Also return false for Armv8.1-M Mainline.
2352 (output_return_instruction): Skip FPSCR clearing if Armv8.1-M
2353 Mainline instructions are available.
2354 (arm_compute_frame_layout): Allocate space in frame for FPCXTNS
2355 when targeting Armv8.1-M Mainline Security Extensions.
2356 (arm_expand_prologue): Save FPCXTNS if this is an Armv8.1-M
2357 Mainline entry function.
2358 (cmse_nonsecure_entry_clear_before_return): Clear IP and r4 if
2359 targeting Armv8.1-M Mainline or successor.
2360 (arm_expand_epilogue): Fix indentation of caller-saved register
2361 clearing. Restore FPCXTNS if this is an Armv8.1-M Mainline
2363 * config/arm/arm.h (TARGET_HAVE_FP_CMSE): New macro.
2364 (FP_SYSREGS): Likewise.
2365 (enum vfp_sysregs_encoding): Define enum.
2366 (fp_sysreg_names): Declare.
2367 * config/arm/unspecs.md (VUNSPEC_VSTR_VLDR): New volatile unspec.
2368 * config/arm/vfp.md (push_fpsysreg_insn): New define_insn.
2369 (pop_fpsysreg_insn): Likewise.
2371 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2372 Thomas Preud'homme <thomas.preudhomme@arm.com>
2374 * config/arm/arm-cpus.in (armv8_1m_main): New feature.
2375 (ARMv4, ARMv4t, ARMv5t, ARMv5te, ARMv5tej, ARMv6, ARMv6j, ARMv6k,
2376 ARMv6z, ARMv6kz, ARMv6zk, ARMv6t2, ARMv6m, ARMv7, ARMv7a, ARMv7ve,
2377 ARMv7r, ARMv7m, ARMv7em, ARMv8a, ARMv8_1a, ARMv8_2a, ARMv8_3a,
2378 ARMv8_4a, ARMv8_5a, ARMv8m_base, ARMv8m_main, ARMv8r): Reindent.
2379 (ARMv8_1m_main): New feature group.
2380 (armv8.1-m.main): New architecture.
2381 * config/arm/arm-tables.opt: Regenerate.
2382 * config/arm/arm.c (arm_arch8_1m_main): Define and default initialize.
2383 (arm_option_reconfigure_globals): Initialize arm_arch8_1m_main.
2384 (arm_options_perform_arch_sanity_checks): Error out when targeting
2385 Armv8.1-M Mainline Security Extensions.
2386 * config/arm/arm.h (arm_arch8_1m_main): Declare.
2388 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2390 * config/aarch64/aarch64-simd-builtins.def (aarch64_bfdot,
2391 aarch64_bfdot_lane, aarch64_bfdot_laneq): New.
2392 * config/aarch64/aarch64-simd.md (aarch64_bfdot, aarch64_bfdot_lane,
2393 aarch64_bfdot_laneq): New.
2394 * config/aarch64/arm_bf16.h (vbfdot_f32, vbfdotq_f32,
2395 vbfdot_lane_f32, vbfdotq_lane_f32, vbfdot_laneq_f32,
2396 vbfdotq_laneq_f32): New.
2397 * config/aarch64/iterators.md (UNSPEC_BFDOT, Vbfdottype,
2398 VBFMLA_W, VBF): New.
2399 (isquadop): Add V4BF, V8BF.
2401 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2403 * config/aarch64/aarch64-builtins.c: (enum aarch64_type_qualifiers):
2404 New qualifier_lane_quadtup_index, TYPES_TERNOP_SSUS,
2405 TYPES_QUADOPSSUS_LANE_QUADTUP, TYPES_QUADOPSSSU_LANE_QUADTUP.
2406 (aarch64_simd_expand_args): Add case SIMD_ARG_LANE_QUADTUP_INDEX.
2407 (aarch64_simd_expand_builtin): Add qualifier_lane_quadtup_index.
2408 * config/aarch64/aarch64-simd-builtins.def (usdot, usdot_lane,
2409 usdot_laneq, sudot_lane,sudot_laneq): New.
2410 * config/aarch64/aarch64-simd.md (aarch64_usdot): New.
2411 (aarch64_<sur>dot_lane): New.
2412 * config/aarch64/arm_neon.h (vusdot_s32): New.
2414 (vusdot_lane_s32): New.
2415 (vsudot_lane_s32): New.
2416 * config/aarch64/iterators.md (DOTPROD_I8MM): New iterator.
2417 (UNSPEC_USDOT, UNSPEC_SUDOT): New unspecs.
2419 2020-01-16 Martin Liska <mliska@suse.cz>
2421 * value-prof.c (dump_histogram_value): Fix
2422 obvious spacing issue.
2424 2020-01-16 Andrew Pinski <apinski@marvell.com>
2426 * tree-ssa-sccvn.c(vn_reference_lookup_3): Check lhs for
2427 !storage_order_barrier_p.
2429 2020-01-16 Andrew Pinski <apinski@marvell.com>
2431 * sched-int.h (_dep): Add unused bit-field field for the padding.
2432 * sched-deps.c (init_dep_1): Init unused field.
2434 2020-01-16 Andrew Pinski <apinski@marvell.com>
2436 * optabs.h (create_expand_operand): Initialize target field also.
2438 2020-01-16 Andre Vieira <andre.simoesdiasvieira@arm.com>
2440 PR tree-optimization/92429
2441 * tree-ssa-loop-niter.h (simplify_replace_tree): Add parameter.
2442 * tree-ssa-loop-niter.c (simplify_replace_tree): Add parameter to
2444 * tree-vect-loop.c (update_epilogue_vinfo): Do not fold when replacing
2447 2020-01-16 Richard Sandiford <richard.sandiford@arm.com>
2449 * config/aarch64/aarch64.c (aarch64_split_sve_subreg_move): Apply
2450 aarch64_sve_int_mode to each mode.
2452 2020-01-15 David Malcolm <dmalcolm@redhat.com>
2454 * doc/analyzer.texi (Overview): Add note about
2455 -fdump-ipa-analyzer.
2457 2020-01-15 Wilco Dijkstra <wdijkstr@arm.com>
2459 PR tree-optimization/93231
2460 * tree-ssa-forwprop.c (optimize_count_trailing_zeroes): Check
2461 input_type is unsigned. Use tree_to_shwi for shift constant.
2462 Check CST_STRING element size is CHAR_TYPE_SIZE bits.
2463 (simplify_count_trailing_zeroes): Add test to handle known non-zero
2464 inputs more efficiently.
2466 2020-01-15 Uroš Bizjak <ubizjak@gmail.com>
2468 * config/i386/i386.md (*movsf_internal): Do not require
2469 SSE2 ISA for alternatives 14 and 15.
2471 2020-01-15 Richard Biener <rguenther@suse.de>
2474 * tree-eh.c (sink_clobbers): If we already visited the destination
2475 block do not defer insertion.
2476 (pass_lower_eh_dispatch::execute): Maintain BB_VISITED for
2477 the purpose of defered insertion.
2479 2020-01-15 Jakub Jelinek <jakub@redhat.com>
2481 * BASE-VER: Bump to 10.0.1.
2483 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
2485 PR tree-optimization/93247
2486 * tree-vect-loop.c (update_epilogue_loop_vinfo): Check the access
2487 type of the stmt that we're going to vectorize.
2489 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
2491 * tree-vect-slp.c (vectorize_slp_instance_root_stmt): Use a
2492 VIEW_CONVERT_EXPR if the vectorized constructor has a diffeent
2495 2020-01-15 Martin Liska <mliska@suse.cz>
2497 * ipa-profile.c (ipa_profile_read_edge_summary): Do not allow
2498 2 calls of streamer_read_hwi in a function call.
2500 2020-01-15 Richard Biener <rguenther@suse.de>
2502 * alias.c (record_alias_subset): Avoid redundant work when
2503 subset is already recorded.
2505 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2507 * doc/invoke.texi (-fdiagnostics-show-cwe): Add note that some of
2508 the analyzer options provide CWE identifiers.
2510 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2512 * tree-diagnostic-path.cc (path_summary::event_range::print):
2513 When testing for UNKNOWN_LOCATION, look through ad-hoc wrappers
2514 using get_pure_location.
2516 2020-01-15 Jakub Jelinek <jakub@redhat.com>
2518 PR tree-optimization/93262
2519 * tree-ssa-dse.c (maybe_trim_memstar_call): For *_chk builtins,
2520 perform head trimming only if the last argument is constant,
2521 either all ones, or larger or equal to head trim, in the latter
2522 case decrease the last argument by head_trim.
2524 PR tree-optimization/93249
2525 * tree-ssa-dse.c: Include builtins.h and gimple-fold.h.
2526 (maybe_trim_memstar_call): Move head_trim and tail_trim vars to
2527 function body scope, reindent. For BUILTIN_IN_STRNCPY*, don't
2528 perform head trim unless we can prove there are no '\0' chars
2529 from the source among the first head_trim chars.
2531 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2533 * Makefile.in (ANALYZER_OBJS): Add analyzer/function-set.o.
2535 2020-01-15 Jakub Jelinek <jakub@redhat.com>
2538 * config/i386/sse.md
2539 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1,
2540 *<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1,
2541 *<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1,
2542 *<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1): Use
2543 just a single alternative instead of two, make operands 1 and 2
2546 2020-01-14 Jan Hubicka <hubicka@ucw.cz>
2549 * ipa-devirt.c (odr_types_equivalent_p): Compare TREE_ADDRESSABLE and
2552 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2554 * Makefile.in (lang_opt_files): Add analyzer.opt.
2555 (ANALYZER_OBJS): New.
2556 (OBJS): Add digraph.o, graphviz.o, ordered-hash-map-tests.o,
2557 tristate.o and ANALYZER_OBJS.
2558 (TEXI_GCCINT_FILES): Add analyzer.texi.
2559 * common.opt (-fanalyzer): New driver option.
2560 * config.in: Regenerate.
2561 * configure: Regenerate.
2562 * configure.ac (--disable-analyzer, ENABLE_ANALYZER): New option.
2563 (gccdepdir): Also create depdir for "analyzer" subdir.
2564 * digraph.cc: New file.
2565 * digraph.h: New file.
2566 * doc/analyzer.texi: New file.
2567 * doc/gccint.texi ("Static Analyzer") New menu item.
2568 (analyzer.texi): Include it.
2569 * doc/invoke.texi ("Static Analyzer Options"): New list and new section.
2570 ("Warning Options"): Add static analysis warnings to the list.
2571 (-Wno-analyzer-double-fclose): New option.
2572 (-Wno-analyzer-double-free): New option.
2573 (-Wno-analyzer-exposure-through-output-file): New option.
2574 (-Wno-analyzer-file-leak): New option.
2575 (-Wno-analyzer-free-of-non-heap): New option.
2576 (-Wno-analyzer-malloc-leak): New option.
2577 (-Wno-analyzer-possible-null-argument): New option.
2578 (-Wno-analyzer-possible-null-dereference): New option.
2579 (-Wno-analyzer-null-argument): New option.
2580 (-Wno-analyzer-null-dereference): New option.
2581 (-Wno-analyzer-stale-setjmp-buffer): New option.
2582 (-Wno-analyzer-tainted-array-index): New option.
2583 (-Wno-analyzer-use-after-free): New option.
2584 (-Wno-analyzer-use-of-pointer-in-stale-stack-frame): New option.
2585 (-Wno-analyzer-use-of-uninitialized-value): New option.
2586 (-Wanalyzer-too-complex): New option.
2587 (-fanalyzer-call-summaries): New warning.
2588 (-fanalyzer-checker=): New warning.
2589 (-fanalyzer-fine-grained): New warning.
2590 (-fno-analyzer-state-merge): New warning.
2591 (-fno-analyzer-state-purge): New warning.
2592 (-fanalyzer-transitivity): New warning.
2593 (-fanalyzer-verbose-edges): New warning.
2594 (-fanalyzer-verbose-state-changes): New warning.
2595 (-fanalyzer-verbosity=): New warning.
2596 (-fdump-analyzer): New warning.
2597 (-fdump-analyzer-callgraph): New warning.
2598 (-fdump-analyzer-exploded-graph): New warning.
2599 (-fdump-analyzer-exploded-nodes): New warning.
2600 (-fdump-analyzer-exploded-nodes-2): New warning.
2601 (-fdump-analyzer-exploded-nodes-3): New warning.
2602 (-fdump-analyzer-supergraph): New warning.
2603 * doc/sourcebuild.texi (dg-require-dot): New.
2604 (dg-check-dot): New.
2605 * gdbinit.in (break-on-saved-diagnostic): New command.
2606 * graphviz.cc: New file.
2607 * graphviz.h: New file.
2608 * ordered-hash-map-tests.cc: New file.
2609 * ordered-hash-map.h: New file.
2610 * passes.def (pass_analyzer): Add before
2611 pass_ipa_whole_program_visibility.
2612 * selftest-run-tests.c (selftest::run_tests): Call
2613 selftest::ordered_hash_map_tests_cc_tests.
2614 * selftest.h (selftest::ordered_hash_map_tests_cc_tests): New
2616 * shortest-paths.h: New file.
2617 * timevar.def (TV_ANALYZER): New timevar.
2618 (TV_ANALYZER_SUPERGRAPH): Likewise.
2619 (TV_ANALYZER_STATE_PURGE): Likewise.
2620 (TV_ANALYZER_PLAN): Likewise.
2621 (TV_ANALYZER_SCC): Likewise.
2622 (TV_ANALYZER_WORKLIST): Likewise.
2623 (TV_ANALYZER_DUMP): Likewise.
2624 (TV_ANALYZER_DIAGNOSTICS): Likewise.
2625 (TV_ANALYZER_SHORTEST_PATHS): Likewise.
2626 * tree-pass.h (make_pass_analyzer): New decl.
2627 * tristate.cc: New file.
2628 * tristate.h: New file.
2630 2020-01-14 Uroš Bizjak <ubizjak@gmail.com>
2633 * config/i386/i386.md (*movsf_internal): Require SSE2 ISA for
2634 alternatives 9 and 10.
2636 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2638 * attribs.c (excl_hash_traits::empty_zero_p): New static constant.
2639 * gcov.c (function_start_pair_hash::empty_zero_p): Likewise.
2640 * graphite.c (struct sese_scev_hash::empty_zero_p): Likewise.
2641 * hash-map-tests.c (selftest::test_nonzero_empty_key): New selftest.
2642 (selftest::hash_map_tests_c_tests): Call it.
2643 * hash-map-traits.h (simple_hashmap_traits::empty_zero_p):
2644 New static constant, using the value of = H::empty_zero_p.
2645 (unbounded_hashmap_traits::empty_zero_p): Likewise, using the value
2646 from default_hash_traits <Value>.
2647 * hash-map.h (hash_map::empty_zero_p): Likewise, using the value
2649 * hash-set-tests.c (value_hash_traits::empty_zero_p): Likewise.
2650 * hash-table.h (hash_table::alloc_entries): Guard the loop of
2651 calls to mark_empty with !Descriptor::empty_zero_p.
2652 (hash_table::empty_slow): Conditionalize the memset call with a
2653 check that Descriptor::empty_zero_p; otherwise, loop through the
2654 entries calling mark_empty on them.
2655 * hash-traits.h (int_hash::empty_zero_p): New static constant.
2656 (pointer_hash::empty_zero_p): Likewise.
2657 (pair_hash::empty_zero_p): Likewise.
2658 * ipa-devirt.c (default_hash_traits <type_pair>::empty_zero_p):
2660 * ipa-prop.c (ipa_bit_ggc_hash_traits::empty_zero_p): Likewise.
2661 (ipa_vr_ggc_hash_traits::empty_zero_p): Likewise.
2662 * profile.c (location_triplet_hash::empty_zero_p): Likewise.
2663 * sanopt.c (sanopt_tree_triplet_hash::empty_zero_p): Likewise.
2664 (sanopt_tree_couple_hash::empty_zero_p): Likewise.
2665 * tree-hasher.h (int_tree_hasher::empty_zero_p): Likewise.
2666 * tree-ssa-sccvn.c (vn_ssa_aux_hasher::empty_zero_p): Likewise.
2667 * tree-vect-slp.c (bst_traits::empty_zero_p): Likewise.
2669 (default_hash_traits<scalar_cond_masked_key>::empty_zero_p):
2672 2020-01-14 Kewen Lin <linkw@gcc.gnu.org>
2674 * cfgloopanal.c (average_num_loop_insns): Free bbs when early return,
2675 fix typo on return value.
2677 2020-01-14 Xiong Hu Luo <luoxhu@linux.ibm.com>
2680 * cgraph.c (symbol_table::create_edge): Init speculative_id and
2682 (cgraph_edge::make_speculative): Add param for setting speculative_id
2684 (cgraph_edge::speculative_call_info): Update comments and find reference
2685 by speculative_id for multiple indirect targets.
2686 (cgraph_edge::resolve_speculation): Decrease the speculations
2687 for indirect edge, drop it's speculative if not direct target
2688 left. Update comments.
2689 (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
2690 (cgraph_node::dump): Print num_speculative_call_targets.
2691 (cgraph_node::verify_node): Don't report error if speculative
2692 edge not include statement.
2693 (cgraph_edge::num_speculative_call_targets_p): New function.
2694 * cgraph.h (int common_target_id): Remove.
2695 (int common_target_probability): Remove.
2696 (num_speculative_call_targets): New variable.
2697 (make_speculative): Add param for setting speculative_id.
2698 (cgraph_edge::num_speculative_call_targets_p): New declare.
2699 (target_prob): New variable.
2700 (speculative_id): New variable.
2701 * ipa-fnsummary.c (analyze_function_body): Create and duplicate
2702 call summaries for multiple speculative call targets.
2703 * cgraphclones.c (cgraph_node::create_clone): Clone speculative_id.
2704 * ipa-profile.c (struct speculative_call_target): New struct.
2705 (class speculative_call_summary): New class.
2706 (class speculative_call_summaries): New class.
2707 (call_sums): New variable.
2708 (ipa_profile_generate_summary): Generate indirect multiple targets summaries.
2709 (ipa_profile_write_edge_summary): New function.
2710 (ipa_profile_write_summary): Stream out indirect multiple targets summaries.
2711 (ipa_profile_dump_all_summaries): New function.
2712 (ipa_profile_read_edge_summary): New function.
2713 (ipa_profile_read_summary_section): New function.
2714 (ipa_profile_read_summary): Stream in indirect multiple targets summaries.
2715 (ipa_profile): Generate num_speculative_call_targets from
2717 * ipa-ref.h (speculative_id): New variable.
2718 * ipa-utils.c (ipa_merge_profiles): Update with target_prob.
2719 * lto-cgraph.c (lto_output_edge): Remove indirect common_target_id and
2720 common_target_probability. Stream out speculative_id and
2721 num_speculative_call_targets.
2722 (input_edge): Likewise.
2723 * predict.c (dump_prediction): Remove edges count assert to be
2725 * symtab.c (symtab_node::create_reference): Init speculative_id.
2726 (symtab_node::clone_references): Clone speculative_id.
2727 (symtab_node::clone_referring): Clone speculative_id.
2728 (symtab_node::clone_reference): Clone speculative_id.
2729 (symtab_node::clear_stmts_in_references): Clear speculative_id.
2730 * tree-inline.c (copy_bb): Duplicate all the speculative edges
2731 if indirect call contains multiple speculative targets.
2732 * value-prof.h (check_ic_target): Remove.
2733 * value-prof.c (gimple_value_profile_transformations):
2734 Use void function gimple_ic_transform.
2735 * value-prof.c (gimple_ic_transform): Handle topn case.
2736 Fix comment typos. Change it to a void function.
2738 2020-01-13 Andrew Pinski <apinski@marvell.com>
2740 * config/aarch64/aarch64-cores.def (octeontx2): New define.
2741 (octeontx2t98): New define.
2742 (octeontx2t96): New define.
2743 (octeontx2t93): New define.
2744 (octeontx2f95): New define.
2745 (octeontx2f95n): New define.
2746 (octeontx2f95mm): New define.
2747 * config/aarch64/aarch64-tune.md: Regenerate.
2748 * doc/invoke.texi (-mcpu=): Document the new cpu types.
2750 2020-01-13 Jason Merrill <jason@redhat.com>
2752 PR c++/33799 - destroy return value if local cleanup throws.
2753 * gimplify.c (gimplify_return_expr): Handle COMPOUND_EXPR.
2755 2020-01-13 Martin Liska <mliska@suse.cz>
2757 * ipa-cp.c (get_max_overall_size): Use newly
2758 renamed param param_ipa_cp_unit_growth.
2759 * params.opt: Remove legacy param name.
2761 2020-01-13 Martin Sebor <msebor@redhat.com>
2763 PR tree-optimization/93213
2764 * tree-ssa-strlen.c (handle_store): Only allow single-byte nul-over-nul
2765 stores to be eliminated.
2767 2020-01-13 Martin Liska <mliska@suse.cz>
2769 * opts.c (print_help): Do not print CL_PARAM
2770 and CL_WARNING for CL_OPTIMIZATION.
2772 2020-01-13 Jonathan Wakely <jwakely@redhat.com>
2775 * doc/invoke.texi (Warning Options): Add caveat about some warnings
2776 depending on optimization settings.
2778 2020-01-13 Jakub Jelinek <jakub@redhat.com>
2780 PR tree-optimization/90838
2781 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
2782 SCALAR_INT_TYPE_MODE directly in CTZ_DEFINED_VALUE_AT_ZERO macro
2783 argument rather than to initialize temporary for targets that
2784 don't use the mode argument at all. Initialize ctzval to avoid
2787 2020-01-10 Thomas Schwinge <thomas@codesourcery.com>
2789 * tree.h (OMP_CLAUSE_USE_DEVICE_PTR_IF_PRESENT): New definition.
2790 * tree-core.h: Document it.
2791 * gimplify.c (gimplify_omp_workshare): Set it.
2792 * omp-low.c (lower_omp_target): Use it.
2793 * tree-pretty-print.c (dump_omp_clause): Print it.
2795 * omp-low.c (lower_omp_target) <OMP_CLAUSE_USE_DEVICE_PTR etc.>:
2796 Assert that for OpenACC we always have 'GOMP_MAP_USE_DEVICE_PTR'.
2798 2020-01-10 David Malcolm <dmalcolm@redhat.com>
2800 * Makefile.in (OBJS): Add tree-diagnostic-path.o.
2801 * common.opt (fdiagnostics-path-format=): New option.
2802 (diagnostic_path_format): New enum.
2803 (fdiagnostics-show-path-depths): New option.
2804 * coretypes.h (diagnostic_event_id_t): New forward decl.
2805 * diagnostic-color.c (color_dict): Add "path".
2806 * diagnostic-event-id.h: New file.
2807 * diagnostic-format-json.cc (json_from_expanded_location): Make
2809 (json_end_diagnostic): Call context->make_json_for_path if it
2810 exists and the diagnostic has a path.
2811 (diagnostic_output_format_init): Clear context->print_path.
2812 * diagnostic-path.h: New file.
2813 * diagnostic-show-locus.c (colorizer::set_range): Special-case
2814 when printing a run of events in a diagnostic_path so that they
2815 all get the same color.
2816 (layout::m_diagnostic_path_p): New field.
2817 (layout::layout): Initialize it.
2818 (layout::print_any_labels): Don't colorize the label text for an
2819 event in a diagnostic_path.
2820 (gcc_rich_location::add_location_if_nearby): Add
2821 "restrict_to_current_line_spans" and "label" params. Pass the
2822 former to layout.maybe_add_location_range; pass the latter
2823 when calling add_range.
2824 * diagnostic.c: Include "diagnostic-path.h".
2825 (diagnostic_initialize): Initialize context->path_format and
2826 context->show_path_depths.
2827 (diagnostic_show_any_path): New function.
2828 (diagnostic_path::interprocedural_p): New function.
2829 (diagnostic_report_diagnostic): Call diagnostic_show_any_path.
2830 (simple_diagnostic_path::num_events): New function.
2831 (simple_diagnostic_path::get_event): New function.
2832 (simple_diagnostic_path::add_event): New function.
2833 (simple_diagnostic_event::simple_diagnostic_event): New ctor.
2834 (simple_diagnostic_event::~simple_diagnostic_event): New dtor.
2835 (debug): New overload taking a diagnostic_path *.
2836 * diagnostic.def (DK_DIAGNOSTIC_PATH): New.
2837 * diagnostic.h (enum diagnostic_path_format): New enum.
2838 (json::value): New forward decl.
2839 (diagnostic_context::path_format): New field.
2840 (diagnostic_context::show_path_depths): New field.
2841 (diagnostic_context::print_path): New callback field.
2842 (diagnostic_context::make_json_for_path): New callback field.
2843 (diagnostic_show_any_path): New decl.
2844 (json_from_expanded_location): New decl.
2845 * doc/invoke.texi (-fdiagnostics-path-format=): New option.
2846 (-fdiagnostics-show-path-depths): New option.
2847 (-fdiagnostics-color): Add "path" to description of default
2848 GCC_COLORS; describe it.
2849 (-fdiagnostics-format=json): Document how diagnostic paths are
2850 represented in the JSON output format.
2851 * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
2852 Add optional params "restrict_to_current_line_spans" and "label".
2853 * opts.c (common_handle_option): Handle
2854 OPT_fdiagnostics_path_format_ and
2855 OPT_fdiagnostics_show_path_depths.
2856 * pretty-print.c: Include "diagnostic-event-id.h".
2857 (pp_format): Implement "%@" format code for printing
2858 diagnostic_event_id_t *.
2859 (selftest::test_pp_format): Add tests for "%@".
2860 * selftest-run-tests.c (selftest::run_tests): Call
2861 selftest::tree_diagnostic_path_cc_tests.
2862 * selftest.h (selftest::tree_diagnostic_path_cc_tests): New decl.
2863 * toplev.c (general_init): Initialize global_dc->path_format and
2864 global_dc->show_path_depths.
2865 * tree-diagnostic-path.cc: New file.
2866 * tree-diagnostic.c (maybe_unwind_expanded_macro_loc): Make
2867 non-static. Drop "diagnostic" param in favor of storing the
2868 original value of "where" and re-using it.
2869 (virt_loc_aware_diagnostic_finalizer): Update for dropped param of
2870 maybe_unwind_expanded_macro_loc.
2871 (tree_diagnostics_defaults): Initialize context->print_path and
2872 context->make_json_for_path.
2873 * tree-diagnostic.h (default_tree_diagnostic_path_printer): New
2875 (default_tree_make_json_for_path): New decl.
2876 (maybe_unwind_expanded_macro_loc): New decl.
2878 2020-01-10 Jakub Jelinek <jakub@redhat.com>
2880 PR tree-optimization/93210
2881 * fold-const.h (native_encode_initializer,
2882 can_native_interpret_type_p): Declare.
2883 * fold-const.c (native_encode_string): Fix up handling with off != -1,
2885 (native_encode_initializer): New function, moved from dwarf2out.c.
2886 Adjust to native_encode_expr compatible arguments, including dry-run
2887 and partial extraction modes. Don't handle STRING_CST.
2888 (can_native_interpret_type_p): No longer static.
2889 * gimple-fold.c (fold_ctor_reference): For native_encode_expr, verify
2890 offset / BITS_PER_UNIT fits into int and don't call it if
2891 can_native_interpret_type_p fails. If suboff is NULL and for
2892 CONSTRUCTOR fold_{,non}array_ctor_reference returns NULL, retry with
2893 native_encode_initializer.
2894 (fold_const_aggregate_ref_1): Formatting fix.
2895 * dwarf2out.c (native_encode_initializer): Moved to fold-const.c.
2896 (tree_add_const_value_attribute): Adjust caller.
2898 PR tree-optimization/90838
2899 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
2900 SCALAR_INT_TYPE_MODE instead of TYPE_MODE as operand of
2901 CTZ_DEFINED_VALUE_AT_ZERO.
2903 2020-01-10 Vladimir Makarov <vmakarov@redhat.com>
2906 * lra-constraints.c (match_reload): Permit input operands have the
2907 same mode as output while other input operands have a different
2910 2020-01-10 Wilco Dijkstra <wdijkstr@arm.com>
2912 PR tree-optimization/90838
2913 * tree-ssa-forwprop.c (check_ctz_array): Add new function.
2914 (check_ctz_string): Likewise.
2915 (optimize_count_trailing_zeroes): Likewise.
2916 (simplify_count_trailing_zeroes): Likewise.
2917 (pass_forwprop::execute): Try ctz simplification.
2918 * match.pd: Add matching for ctz idioms.
2920 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2922 * config/aarch64/aarch64.c (aarch64_invalid_conversion): New function
2924 (aarch64_invalid_unary_op): New function for target hook.
2925 (aarch64_invalid_binary_op): New function for target hook.
2927 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2929 * config.gcc: Add arm_bf16.h.
2930 * config/aarch64/aarch64-builtins.c
2931 (aarch64_simd_builtin_std_type): Add BFmode.
2932 (aarch64_init_simd_builtin_types): Define element types for vector
2934 (aarch64_init_bf16_types): New function.
2935 (aarch64_general_init_builtins): Add arm_init_bf16_types function call.
2936 * config/aarch64/aarch64-modes.def: Add BFmode and V4BF, V8BF vector
2938 * config/aarch64/aarch64-simd-builtin-types.def: Add BF SIMD types.
2939 * config/aarch64/aarch64-simd.md: Add BF vector types to NEON move
2941 * config/aarch64/aarch64.h (AARCH64_VALID_SIMD_DREG_MODE): Add V4BF.
2942 (AARCH64_VALID_SIMD_QREG_MODE): Add V8BF.
2943 * config/aarch64/aarch64.c
2944 (aarch64_classify_vector_mode): Add support for BF types.
2945 (aarch64_gimplify_va_arg_expr): Add support for BF types.
2946 (aarch64_vq_mode): Add support for BF types.
2947 (aarch64_simd_container_mode): Add support for BF types.
2948 (aarch64_mangle_type): Add support for BF scalar type.
2949 * config/aarch64/aarch64.md: Add BFmode to movhf pattern.
2950 * config/aarch64/arm_bf16.h: New file.
2951 * config/aarch64/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
2952 * config/aarch64/iterators.md: Add BF types to mode attributes.
2953 (HFBF, GPF_TF_F16_MOV, VDMOV, VQMOV, VQMOV_NO2Em VALL_F16MOV): New.
2955 2020-01-10 Jason Merrill <jason@redhat.com>
2957 PR c++/93173 - incorrect tree sharing.
2958 * gimplify.c (copy_if_shared): No longer static.
2959 * gimplify.h: Declare it.
2961 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
2963 * doc/invoke.texi (-msve-vector-bits=): Document that
2964 -msve-vector-bits=128 now generates VL-specific code for
2965 little-endian targets.
2966 * config/aarch64/aarch64-sve-builtins.cc (register_builtin_types): Use
2967 build_vector_type_for_mode to construct the data vector types.
2968 * config/aarch64/aarch64.c (aarch64_convert_sve_vector_bits): Generate
2969 VL-specific code for -msve-vector-bits=128 on little-endian targets.
2970 (aarch64_simd_container_mode): Always prefer Advanced SIMD modes
2971 for 128-bit vectors.
2973 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
2975 * config/aarch64/aarch64.c (aarch64_evpc_sel): Fix gen_vcond_mask
2978 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
2980 * config/aarch64/aarch64-builtins.c
2981 (aarch64_builtin_vectorized_function): Check for specific vector modes,
2982 rather than checking the number of elements and the element mode.
2984 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
2986 * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
2987 get_related_vectype_for_scalar_type rather than build_vector_type
2988 to create the index type for a conditional reduction.
2990 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
2992 * tree-vect-loop.c (update_epilogue_loop_vinfo): Update DR_REF
2993 for any type of gather or scatter, including strided accesses.
2995 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
2997 * tree-vectorizer.h (get_dr_vinfo_offset): Add missing function
3000 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
3002 * tree-vect-data-refs.c (vect_create_addr_base_for_vector_ref): Use
3004 * tree-vect-loop.c (update_epilogue_loop_vinfo): Remove orig_drs_init
3005 parameter and its use to reset DR_OFFSET's.
3006 (vect_transform_loop): Remove orig_drs_init argument.
3007 * tree-vect-loop-manip.c (vect_update_init_of_dr): Update the offset
3008 member of dr_vec_info rather than the offset of the associated
3009 data_reference's innermost_loop_behavior.
3010 (vect_update_init_of_dr): Pass dr_vec_info instead of data_reference.
3011 (vect_do_peeling): Remove orig_drs_init parameter and its construction.
3012 * tree-vect-stmts.c (check_scan_store): Replace use of DR_OFFSET with
3013 get_dr_vinfo_offset.
3014 (vectorizable_store): Likewise.
3015 (vectorizable_load): Likewise.
3017 2020-01-10 Richard Biener <rguenther@suse.de>
3019 * gimple-ssa-store-merging
3020 (pass_store_merging::terminate_all_aliasing_chains): Cache alias info.
3022 2020-01-10 Martin Liska <mliska@suse.cz>
3025 * ipa-inline-analysis.c (offline_size): Make proper parenthesis
3026 encapsulation that was there before r280040.
3028 2020-01-10 Richard Biener <rguenther@suse.de>
3031 * tree-eh.c (sink_clobbers): Move clobbers to out-of-IL
3032 sequences to avoid walking them again for secondary opportunities.
3033 (pass_lower_eh_dispatch::execute): Instead actually insert
3036 2020-01-10 Richard Biener <rguenther@suse.de>
3039 * tree-eh.c (redirect_eh_edge_1): Avoid some work if possible.
3040 (cleanup_all_empty_eh): Walk landing pads in reverse order to
3041 avoid quadraticness.
3043 2020-01-10 Martin Jambor <mjambor@suse.cz>
3045 * params.opt (param_ipa_sra_max_replacements): Mark as Optimization.
3046 * ipa-sra.c (pull_accesses_from_callee): New parameter caller, use it
3047 to get param_ipa_sra_max_replacements.
3048 (param_splitting_across_edge): Pass the caller to
3049 pull_accesses_from_callee.
3051 2020-01-10 Martin Jambor <mjambor@suse.cz>
3053 * params.opt (param_ipcp_unit_growth): Mark as Optimization.
3054 * ipa-cp.c (max_new_size): Removed.
3055 (orig_overall_size): New variable.
3056 (get_max_overall_size): New function.
3057 (estimate_local_effects): Use it. Adjust dump.
3058 (decide_about_value): Likewise.
3059 (ipcp_propagate_stage): Do not calculate max_new_size, just store
3060 orig_overall_size. Adjust dump.
3061 (ipa_cp_c_finalize): Clear orig_overall_size instead of max_new_size.
3063 2020-01-10 Martin Jambor <mjambor@suse.cz>
3065 * params.opt (param_ipa_max_agg_items): Mark as Optimization
3066 * ipa-cp.c (merge_agg_lats_step): New parameter max_agg_items, use
3067 instead of param_ipa_max_agg_items.
3068 (merge_aggregate_lattices): Extract param_ipa_max_agg_items from
3069 optimization info for the callee.
3071 2020-01-09 Kwok Cheung Yeung <kcy@codesourcery.com>
3073 * lto-streamer-in.c (input_function): Remove streamed-in inline debug
3074 markers if debug_inline_points is false.
3076 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3078 * config.gcc (aarch64*-*-*): Add aarch64-sve-builtins-sve2.o to
3080 * config/aarch64/t-aarch64 (aarch64-sve-builtins.o): Depend on
3081 aarch64-sve-builtins-base.def, aarch64-sve-builtins-sve2.def and
3082 aarch64-sve-builtins-sve2.h.
3083 (aarch64-sve-builtins-sve2.o): New rule.
3084 * config/aarch64/aarch64.h (AARCH64_ISA_SVE2_AES): New macro.
3085 (AARCH64_ISA_SVE2_BITPERM, AARCH64_ISA_SVE2_SHA3): Likewise.
3086 (AARCH64_ISA_SVE2_SM4, TARGET_SVE2_AES, TARGET_SVE2_BITPERM): Likewise.
3087 (TARGET_SVE2_SHA, TARGET_SVE2_SM4): Likewise.
3088 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
3089 TARGET_SVE2_AES, TARGET_SVE2_BITPERM, TARGET_SVE2_SHA3 and
3091 * config/aarch64/aarch64-sve.md: Update comments with SVE2
3092 instructions that are handled here.
3093 (@cond_asrd<mode>): Generalize to...
3094 (@cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>): ...this.
3095 (*cond_asrd<mode>_2): Generalize to...
3096 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_2): ...this.
3097 (*cond_asrd<mode>_z): Generalize to...
3098 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_z): ...this.
3099 * config/aarch64/aarch64.md (UNSPEC_LDNT1_GATHER): New unspec.
3100 (UNSPEC_STNT1_SCATTER, UNSPEC_WHILEGE, UNSPEC_WHILEGT): Likewise.
3101 (UNSPEC_WHILEHI, UNSPEC_WHILEHS): Likewise.
3102 * config/aarch64/aarch64-sve2.md (@aarch64_gather_ldnt<mode>): New
3104 (@aarch64_gather_ldnt_<ANY_EXTEND:optab><SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
3105 (@aarch64_scatter_stnt<mode>): Likewise.
3106 (@aarch64_scatter_stnt_<SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
3107 (@aarch64_mul_lane_<mode>): Likewise.
3108 (@aarch64_sve_suqadd<mode>_const): Likewise.
3109 (*<sur>h<addsub><mode>): Generalize to...
3110 (@aarch64_pred_<SVE2_COND_INT_BINARY_REV:sve_int_op><mode>): ...this
3112 (@cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>): New expander.
3113 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_2): New pattern.
3114 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_3): Likewise.
3115 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_any): Likewise.
3116 (*cond_<SVE2_COND_INT_BINARY_NOREV:sve_int_op><mode>_z): Likewise.
3117 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op><mode>):: Likewise.
3118 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op>_lane_<mode>): Likewise.
3119 (@aarch64_pred_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): Likewise.
3120 (@cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): New expander.
3121 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_2): New pattern.
3122 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_3): Likewise.
3123 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_any): Likewise.
3124 (@aarch64_sve_<SVE2_INT_TERNARY:sve_int_op><mode>): Likewise.
3125 (@aarch64_sve_<SVE2_INT_TERNARY_LANE:sve_int_op>_lane_<mode>)
3126 (@aarch64_sve_add_mul_lane_<mode>): Likewise.
3127 (@aarch64_sve_sub_mul_lane_<mode>): Likewise.
3128 (@aarch64_sve2_xar<mode>): Likewise.
3129 (@aarch64_sve2_bcax<mode>): Likewise.
3130 (*aarch64_sve2_eor3<mode>): Rename to...
3131 (@aarch64_sve2_eor3<mode>): ...this.
3132 (@aarch64_sve2_bsl<mode>): New expander.
3133 (@aarch64_sve2_nbsl<mode>): Likewise.
3134 (@aarch64_sve2_bsl1n<mode>): Likewise.
3135 (@aarch64_sve2_bsl2n<mode>): Likewise.
3136 (@aarch64_sve_add_<SHIFTRT:sve_int_op><mode>): Likewise.
3137 (*aarch64_sve2_sra<mode>): Add MOVPRFX support.
3138 (@aarch64_sve_add_<VRSHR_N:sve_int_op><mode>): New pattern.
3139 (@aarch64_sve_<SVE2_INT_SHIFT_INSERT:sve_int_op><mode>): Likewise.
3140 (@aarch64_sve2_<USMAX:su>aba<mode>): New expander.
3141 (*aarch64_sve2_<USMAX:su>aba<mode>): New pattern.
3142 (@aarch64_sve_<SVE2_INT_BINARY_WIDE:sve_int_op><mode>): Likewise.
3143 (<su>mull<bt><Vwide>): Generalize to...
3144 (@aarch64_sve_<SVE2_INT_BINARY_LONG:sve_int_op><mode>): ...this new
3146 (@aarch64_sve_<SVE2_INT_BINARY_LONG_lANE:sve_int_op>_lane_<mode>)
3147 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_LONG:sve_int_op><mode>)
3148 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG:sve_int_op><mode>)
3149 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3150 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG:sve_int_op><mode>)
3151 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3152 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG:sve_int_op><mode>)
3153 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3154 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG:sve_int_op><mode>)
3155 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3156 (@aarch64_sve_<SVE2_FP_TERNARY_LONG:sve_fp_op><mode>): New patterns.
3157 (@aarch64_<SVE2_FP_TERNARY_LONG_LANE:sve_fp_op>_lane_<mode>)
3158 (@aarch64_sve_<SVE2_INT_UNARY_NARROWB:sve_int_op><mode>): Likewise.
3159 (@aarch64_sve_<SVE2_INT_UNARY_NARROWT:sve_int_op><mode>): Likewise.
3160 (@aarch64_sve_<SVE2_INT_BINARY_NARROWB:sve_int_op><mode>): Likewise.
3161 (@aarch64_sve_<SVE2_INT_BINARY_NARROWT:sve_int_op><mode>): Likewise.
3162 (<SHRNB:r>shrnb<mode>): Generalize to...
3163 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWB:sve_int_op><mode>): ...this
3165 (<SHRNT:r>shrnt<mode>): Generalize to...
3166 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWT:sve_int_op><mode>): ...this
3168 (@aarch64_pred_<SVE2_INT_BINARY_PAIR:sve_int_op><mode>): New pattern.
3169 (@aarch64_pred_<SVE2_FP_BINARY_PAIR:sve_fp_op><mode>): Likewise.
3170 (@cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>): New expander.
3171 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_2): New pattern.
3172 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_z): Likewise.
3173 (@aarch64_sve_<SVE2_INT_CADD:optab><mode>): Likewise.
3174 (@aarch64_sve_<SVE2_INT_CMLA:optab><mode>): Likewise.
3175 (@aarch64_<SVE2_INT_CMLA:optab>_lane_<mode>): Likewise.
3176 (@aarch64_sve_<SVE2_INT_CDOT:optab><mode>): Likewise.
3177 (@aarch64_<SVE2_INT_CDOT:optab>_lane_<mode>): Likewise.
3178 (@aarch64_pred_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): Likewise.
3179 (@cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New expander.
3180 (*cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New pattern.
3181 (@aarch64_sve2_cvtnt<mode>): Likewise.
3182 (@aarch64_pred_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): Likewise.
3183 (@cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): New expander.
3184 (*cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>_any): New pattern.
3185 (@aarch64_sve2_cvtxnt<mode>): Likewise.
3186 (@aarch64_pred_<SVE2_U32_UNARY:sve_int_op><mode>): Likewise.
3187 (@cond_<SVE2_U32_UNARY:sve_int_op><mode>): New expander.
3188 (*cond_<SVE2_U32_UNARY:sve_int_op><mode>): New pattern.
3189 (@aarch64_pred_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): Likewise.
3190 (@cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New expander.
3191 (*cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New pattern.
3192 (@aarch64_sve2_pmul<mode>): Likewise.
3193 (@aarch64_sve_<SVE2_PMULL:optab><mode>): Likewise.
3194 (@aarch64_sve_<SVE2_PMULL_PAIR:optab><mode>): Likewise.
3195 (@aarch64_sve2_tbl2<mode>): Likewise.
3196 (@aarch64_sve2_tbx<mode>): Likewise.
3197 (@aarch64_sve_<SVE2_INT_BITPERM:sve_int_op><mode>): Likewise.
3198 (@aarch64_sve2_histcnt<mode>): Likewise.
3199 (@aarch64_sve2_histseg<mode>): Likewise.
3200 (@aarch64_pred_<SVE2_MATCH:sve_int_op><mode>): Likewise.
3201 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_cc): Likewise.
3202 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_ptest): Likewise.
3203 (aarch64_sve2_aes<CRYPTO_AES:aes_op>): Likewise.
3204 (aarch64_sve2_aes<CRYPTO_AESMC:aesmc_op>): Likewise.
3205 (*aarch64_sve2_aese_fused, *aarch64_sve2_aesd_fused): Likewise.
3206 (aarch64_sve2_rax1, aarch64_sve2_sm4e, aarch64_sve2_sm4ekey): Likewise.
3207 (<su>mulh<r>s<mode>3): Update after above pattern name changes.
3208 * config/aarch64/iterators.md (VNx16QI_ONLY, VNx4SF_ONLY)
3209 (SVE_STRUCT2, SVE_FULL_BHI, SVE_FULL_HSI, SVE_FULL_HDI)
3210 (SVE2_PMULL_PAIR_I): New mode iterators.
3211 (UNSPEC_ADCLB, UNSPEC_ADCLT, UNSPEC_ADDHNB, UNSPEC_ADDHNT, UNSPEC_BDEP)
3212 (UNSPEC_BEXT, UNSPEC_BGRP, UNSPEC_CADD90, UNSPEC_CADD270, UNSPEC_CDOT)
3213 (UNSPEC_CDOT90, UNSPEC_CDOT180, UNSPEC_CDOT270, UNSPEC_CMLA)
3214 (UNSPEC_CMLA90, UNSPEC_CMLA180, UNSPEC_CMLA270, UNSPEC_COND_FCVTLT)
3215 (UNSPEC_COND_FCVTNT, UNSPEC_COND_FCVTX, UNSPEC_COND_FCVTXNT)
3216 (UNSPEC_COND_FLOGB, UNSPEC_EORBT, UNSPEC_EORTB, UNSPEC_FADDP)
3217 (UNSPEC_FMAXP, UNSPEC_FMAXNMP, UNSPEC_FMLALB, UNSPEC_FMLALT)
3218 (UNSPEC_FMLSLB, UNSPEC_FMLSLT, UNSPEC_FMINP, UNSPEC_FMINNMP)
3219 (UNSPEC_HISTCNT, UNSPEC_HISTSEG, UNSPEC_MATCH, UNSPEC_NMATCH)
3220 (UNSPEC_PMULLB, UNSPEC_PMULLB_PAIR, UNSPEC_PMULLT, UNSPEC_PMULLT_PAIR)
3221 (UNSPEC_RADDHNB, UNSPEC_RADDHNT, UNSPEC_RSUBHNB, UNSPEC_RSUBHNT)
3222 (UNSPEC_SLI, UNSPEC_SRI, UNSPEC_SABDLB, UNSPEC_SABDLT, UNSPEC_SADDLB)
3223 (UNSPEC_SADDLBT, UNSPEC_SADDLT, UNSPEC_SADDWB, UNSPEC_SADDWT)
3224 (UNSPEC_SBCLB, UNSPEC_SBCLT, UNSPEC_SMAXP, UNSPEC_SMINP)
3225 (UNSPEC_SQCADD90, UNSPEC_SQCADD270, UNSPEC_SQDMULLB, UNSPEC_SQDMULLBT)
3226 (UNSPEC_SQDMULLT, UNSPEC_SQRDCMLAH, UNSPEC_SQRDCMLAH90)
3227 (UNSPEC_SQRDCMLAH180, UNSPEC_SQRDCMLAH270, UNSPEC_SQRSHRNB)
3228 (UNSPEC_SQRSHRNT, UNSPEC_SQRSHRUNB, UNSPEC_SQRSHRUNT, UNSPEC_SQSHRNB)
3229 (UNSPEC_SQSHRNT, UNSPEC_SQSHRUNB, UNSPEC_SQSHRUNT, UNSPEC_SQXTNB)
3230 (UNSPEC_SQXTNT, UNSPEC_SQXTUNB, UNSPEC_SQXTUNT, UNSPEC_SSHLLB)
3231 (UNSPEC_SSHLLT, UNSPEC_SSUBLB, UNSPEC_SSUBLBT, UNSPEC_SSUBLT)
3232 (UNSPEC_SSUBLTB, UNSPEC_SSUBWB, UNSPEC_SSUBWT, UNSPEC_SUBHNB)
3233 (UNSPEC_SUBHNT, UNSPEC_TBL2, UNSPEC_UABDLB, UNSPEC_UABDLT)
3234 (UNSPEC_UADDLB, UNSPEC_UADDLT, UNSPEC_UADDWB, UNSPEC_UADDWT)
3235 (UNSPEC_UMAXP, UNSPEC_UMINP, UNSPEC_UQRSHRNB, UNSPEC_UQRSHRNT)
3236 (UNSPEC_UQSHRNB, UNSPEC_UQSHRNT, UNSPEC_UQXTNB, UNSPEC_UQXTNT)
3237 (UNSPEC_USHLLB, UNSPEC_USHLLT, UNSPEC_USUBLB, UNSPEC_USUBLT)
3238 (UNSPEC_USUBWB, UNSPEC_USUBWT): New unspecs.
3239 (UNSPEC_SMULLB, UNSPEC_SMULLT, UNSPEC_UMULLB, UNSPEC_UMULLT)
3240 (UNSPEC_SMULHS, UNSPEC_SMULHRS, UNSPEC_UMULHS, UNSPEC_UMULHRS)
3241 (UNSPEC_RSHRNB, UNSPEC_RSHRNT, UNSPEC_SHRNB, UNSPEC_SHRNT): Move
3243 (VNARROW, Ventype): New mode attributes.
3244 (Vewtype): Handle VNx2DI. Fix typo in comment.
3245 (VDOUBLE): New mode attribute.
3246 (sve_lane_con): Handle VNx8HI.
3247 (SVE_INT_UNARY): Include ss_abs and ss_neg for TARGET_SVE2.
3248 (SVE_INT_BINARY): Likewise ss_plus, us_plus, ss_minus and us_minus.
3249 (sve_int_op, sve_int_op_rev): Handle the above codes.
3250 (sve_pred_int_rhs2_operand): Likewise.
3251 (MULLBT, SHRNB, SHRNT): Delete.
3252 (SVE_INT_SHIFT_IMM): New int iterator.
3253 (SVE_WHILE): Add UNSPEC_WHILEGE, UNSPEC_WHILEGT, UNSPEC_WHILEHI
3254 and UNSPEC_WHILEHS for TARGET_SVE2.
3255 (SVE2_U32_UNARY, SVE2_INT_UNARY_NARROWB, SVE2_INT_UNARY_NARROWT)
3256 (SVE2_INT_BINARY, SVE2_INT_BINARY_LANE, SVE2_INT_BINARY_LONG)
3257 (SVE2_INT_BINARY_LONG_LANE, SVE2_INT_BINARY_NARROWB)
3258 (SVE2_INT_BINARY_NARROWT, SVE2_INT_BINARY_PAIR, SVE2_FP_BINARY_PAIR)
3259 (SVE2_INT_BINARY_PAIR_LONG, SVE2_INT_BINARY_WIDE): New int iterators.
3260 (SVE2_INT_SHIFT_IMM_LONG, SVE2_INT_SHIFT_IMM_NARROWB): Likewise.
3261 (SVE2_INT_SHIFT_IMM_NARROWT, SVE2_INT_SHIFT_INSERT, SVE2_INT_CADD)
3262 (SVE2_INT_BITPERM, SVE2_INT_TERNARY, SVE2_INT_TERNARY_LANE): Likewise.
3263 (SVE2_FP_TERNARY_LONG, SVE2_FP_TERNARY_LONG_LANE, SVE2_INT_CMLA)
3264 (SVE2_INT_CDOT, SVE2_INT_ADD_BINARY_LONG, SVE2_INT_QADD_BINARY_LONG)
3265 (SVE2_INT_SUB_BINARY_LONG, SVE2_INT_QSUB_BINARY_LONG): Likewise.
3266 (SVE2_INT_ADD_BINARY_LONG_LANE, SVE2_INT_QADD_BINARY_LONG_LANE)
3267 (SVE2_INT_SUB_BINARY_LONG_LANE, SVE2_INT_QSUB_BINARY_LONG_LANE)
3268 (SVE2_COND_INT_UNARY_FP, SVE2_COND_FP_UNARY_LONG): Likewise.
3269 (SVE2_COND_FP_UNARY_NARROWB, SVE2_COND_INT_BINARY): Likewise.
3270 (SVE2_COND_INT_BINARY_NOREV, SVE2_COND_INT_BINARY_REV): Likewise.
3271 (SVE2_COND_INT_SHIFT, SVE2_MATCH, SVE2_PMULL): Likewise.
3272 (optab): Handle the new unspecs.
3273 (su, r): Remove entries for UNSPEC_SHRNB, UNSPEC_SHRNT, UNSPEC_RSHRNB
3275 (lr): Handle the new unspecs.
3277 (cmp_op, while_optab_cmp, sve_int_op): Handle the new unspecs.
3278 (sve_int_op_rev, sve_int_add_op, sve_int_qadd_op, sve_int_sub_op)
3279 (sve_int_qsub_op): New int attributes.
3280 (sve_fp_op, rot): Handle the new unspecs.
3281 * config/aarch64/aarch64-sve-builtins.h
3282 (function_resolver::require_matching_pointer_type): Declare.
3283 (function_resolver::resolve_unary): Add an optional boolean argument.
3284 (function_resolver::finish_opt_n_resolution): Add an optional
3285 type_suffix_index argument.
3286 (gimple_folder::redirect_call): Declare.
3287 (gimple_expander::prepare_gather_address_operands): Add an optional
3289 * config/aarch64/aarch64-sve-builtins.cc: Include
3290 aarch64-sve-builtins-sve2.h.
3291 (TYPES_b_unsigned, TYPES_b_integer, TYPES_bh_integer): New macros.
3292 (TYPES_bs_unsigned, TYPES_hs_signed, TYPES_hs_integer): Likewise.
3293 (TYPES_hd_unsigned, TYPES_hsd_signed): Likewise.
3294 (TYPES_hsd_integer): Use TYPES_hsd_signed.
3295 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): New macros.
3296 (TYPES_s_unsigned): Likewise.
3297 (TYPES_s_integer): Use TYPES_s_unsigned.
3298 (TYPES_sd_signed, TYPES_sd_unsigned): New macros.
3299 (TYPES_sd_integer): Use them.
3300 (TYPES_d_unsigned): New macro.
3301 (TYPES_d_integer): Use it.
3302 (TYPES_d_data, TYPES_cvt_long, TYPES_cvt_narrow_s): New macros.
3303 (TYPES_cvt_narrow): Likewise.
3304 (DEF_SVE_TYPES_ARRAY): Include the new types macros above.
3305 (preds_mx): New variable.
3306 (function_builder::add_overloaded_function): Allow the new feature
3307 set to be more restrictive than the original one.
3308 (function_resolver::infer_pointer_type): Remove qualifiers from
3309 the pointer type before printing it.
3310 (function_resolver::require_matching_pointer_type): New function.
3311 (function_resolver::resolve_sv_displacement): Handle functions
3312 that don't support 32-bit vector indices or svint32_t vector offsets.
3313 (function_resolver::finish_opt_n_resolution): Take the inferred type
3314 as a separate argument.
3315 (function_resolver::resolve_unary): Optionally treat all forms in
3316 the same way as normal merging functions.
3317 (gimple_folder::redirect_call): New function.
3318 (function_expander::prepare_gather_address_operands): Add an argument
3319 that says whether scaled forms are available. If they aren't,
3320 handle scaling of vector indices and don't add the extension and
3322 (function_expander::map_to_unspecs): If aarch64_sve isn't available,
3323 fall back to using cond_* instead.
3324 * config/aarch64/aarch64-sve-builtins-functions.h (rtx_code_function):
3325 Split out the member variables into...
3326 (rtx_code_function_base): ...this new base class.
3327 (rtx_code_function_rotated): Inherit rtx_code_function_base.
3328 (unspec_based_function): Split out the member variables into...
3329 (unspec_based_function_base): ...this new base class.
3330 (unspec_based_function_rotated): Inherit unspec_based_function_base.
3331 (unspec_based_function_exact_insn): New class.
3332 (unspec_based_add_function, unspec_based_add_lane_function)
3333 (unspec_based_lane_function, unspec_based_pred_function)
3334 (unspec_based_qadd_function, unspec_based_qadd_lane_function)
3335 (unspec_based_qsub_function, unspec_based_qsub_lane_function)
3336 (unspec_based_sub_function, unspec_based_sub_lane_function): New
3338 (unspec_based_fused_function): New class.
3339 (unspec_based_mla_function, unspec_based_mls_function): New typedefs.
3340 (unspec_based_fused_lane_function): New class.
3341 (unspec_based_mla_lane_function, unspec_based_mls_lane_function): New
3343 (CODE_FOR_MODE1): New macro.
3344 (fixed_insn_function): New class.
3345 (while_comparison): Likewise.
3346 * config/aarch64/aarch64-sve-builtins-shapes.h (binary_long_lane)
3347 (binary_long_opt_n, binary_narrowb_opt_n, binary_narrowt_opt_n)
3348 (binary_to_uint, binary_wide, binary_wide_opt_n, compare, compare_ptr)
3349 (load_ext_gather_index_restricted, load_ext_gather_offset_restricted)
3350 (load_gather_sv_restricted, shift_left_imm_long): Declare.
3351 (shift_left_imm_to_uint, shift_right_imm_narrowb): Likewise.
3352 (shift_right_imm_narrowt, shift_right_imm_narrowb_to_uint): Likewise.
3353 (shift_right_imm_narrowt_to_uint, store_scatter_index_restricted)
3354 (store_scatter_offset_restricted, tbl_tuple, ternary_long_lane)
3355 (ternary_long_opt_n, ternary_qq_lane_rotate, ternary_qq_rotate)
3356 (ternary_shift_left_imm, ternary_shift_right_imm, ternary_uint)
3357 (unary_convert_narrowt, unary_long, unary_narrowb, unary_narrowt)
3358 (unary_narrowb_to_uint, unary_narrowt_to_uint, unary_to_int): Likewise.
3359 * config/aarch64/aarch64-sve-builtins-shapes.cc (apply_predication):
3360 Also add an initial argument for unary_convert_narrowt, regardless
3361 of the predication type.
3362 (build_32_64): Allow loads and stores to specify MODE_none.
3363 (build_sv_index64, build_sv_uint_offset): New functions.
3364 (long_type_suffix): New function.
3365 (binary_imm_narrowb_base, binary_imm_narrowt_base): New classes.
3366 (binary_imm_long_base, load_gather_sv_base): Likewise.
3367 (shift_right_imm_narrow_wrapper, ternary_shift_imm_base): Likewise.
3368 (ternary_resize2_opt_n_base, ternary_resize2_lane_base): Likewise.
3369 (unary_narrowb_base, unary_narrowt_base): Likewise.
3370 (binary_long_lane_def, binary_long_lane): New shape.
3371 (binary_long_opt_n_def, binary_long_opt_n): Likewise.
3372 (binary_narrowb_opt_n_def, binary_narrowb_opt_n): Likewise.
3373 (binary_narrowt_opt_n_def, binary_narrowt_opt_n): Likewise.
3374 (binary_to_uint_def, binary_to_uint): Likewise.
3375 (binary_wide_def, binary_wide): Likewise.
3376 (binary_wide_opt_n_def, binary_wide_opt_n): Likewise.
3377 (compare_def, compare): Likewise.
3378 (compare_ptr_def, compare_ptr): Likewise.
3379 (load_ext_gather_index_restricted_def,
3380 load_ext_gather_index_restricted): Likewise.
3381 (load_ext_gather_offset_restricted_def,
3382 load_ext_gather_offset_restricted): Likewise.
3383 (load_gather_sv_def): Inherit from load_gather_sv_base.
3384 (load_gather_sv_restricted_def, load_gather_sv_restricted): New shape.
3385 (shift_left_imm_def, shift_left_imm): Likewise.
3386 (shift_left_imm_long_def, shift_left_imm_long): Likewise.
3387 (shift_left_imm_to_uint_def, shift_left_imm_to_uint): Likewise.
3388 (store_scatter_index_restricted_def,
3389 store_scatter_index_restricted): Likewise.
3390 (store_scatter_offset_restricted_def,
3391 store_scatter_offset_restricted): Likewise.
3392 (tbl_tuple_def, tbl_tuple): Likewise.
3393 (ternary_long_lane_def, ternary_long_lane): Likewise.
3394 (ternary_long_opt_n_def, ternary_long_opt_n): Likewise.
3395 (ternary_qq_lane_def): Inherit from ternary_resize2_lane_base.
3396 (ternary_qq_lane_rotate_def, ternary_qq_lane_rotate): New shape
3397 (ternary_qq_opt_n_def): Inherit from ternary_resize2_opt_n_base.
3398 (ternary_qq_rotate_def, ternary_qq_rotate): New shape.
3399 (ternary_shift_left_imm_def, ternary_shift_left_imm): Likewise.
3400 (ternary_shift_right_imm_def, ternary_shift_right_imm): Likewise.
3401 (ternary_uint_def, ternary_uint): Likewise.
3402 (unary_convert): Fix typo in comment.
3403 (unary_convert_narrowt_def, unary_convert_narrowt): New shape.
3404 (unary_long_def, unary_long): Likewise.
3405 (unary_narrowb_def, unary_narrowb): Likewise.
3406 (unary_narrowt_def, unary_narrowt): Likewise.
3407 (unary_narrowb_to_uint_def, unary_narrowb_to_uint): Likewise.
3408 (unary_narrowt_to_uint_def, unary_narrowt_to_uint): Likewise.
3409 (unary_to_int_def, unary_to_int): Likewise.
3410 * config/aarch64/aarch64-sve-builtins-base.cc (unspec_cmla)
3411 (unspec_fcmla, unspec_cond_fcmla, expand_mla_mls_lane): New functions.
3412 (svasrd_impl): Delete.
3413 (svcadd_impl::expand): Handle integer operations too.
3414 (svcmla_impl::expand, svcmla_lane::expand): Likewise, using the
3415 new functions to derive the unspec numbers.
3416 (svmla_svmls_lane_impl): Replace with...
3417 (svmla_lane_impl, svmls_lane_impl): ...these new classes. Handle
3418 integer operations too.
3419 (svwhile_impl): Rename to...
3420 (svwhilelx_impl): ...this and inherit from while_comparison.
3421 (svasrd): Use unspec_based_function.
3422 (svmla_lane): Use svmla_lane_impl.
3423 (svmls_lane): Use svmls_lane_impl.
3424 (svrecpe, svrsqrte): Handle unsigned integer operations too.
3425 (svwhilele, svwhilelt): Use svwhilelx_impl.
3426 * config/aarch64/aarch64-sve-builtins-sve2.h: New file.
3427 * config/aarch64/aarch64-sve-builtins-sve2.cc: Likewise.
3428 * config/aarch64/aarch64-sve-builtins-sve2.def: Likewise.
3429 * config/aarch64/aarch64-sve-builtins.def: Include
3430 aarch64-sve-builtins-sve2.def.
3432 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3434 * config/aarch64/aarch64-protos.h (aarch64_sve_arith_immediate_p)
3435 (aarch64_sve_sqadd_sqsub_immediate_p): Add a machine_mode argument.
3436 * config/aarch64/aarch64.c (aarch64_sve_arith_immediate_p)
3437 (aarch64_sve_sqadd_sqsub_immediate_p): Likewise. Handle scalar
3438 immediates as well as vector ones.
3439 * config/aarch64/predicates.md (aarch64_sve_arith_immediate)
3440 (aarch64_sve_sub_arith_immediate, aarch64_sve_qadd_immediate)
3441 (aarch64_sve_qsub_immediate): Update calls accordingly.
3443 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3445 * config/aarch64/aarch64-sve2.md: Add banner comments.
3446 (<su>mulh<r>s<mode>3): Move further up file.
3447 (<su>mull<bt><Vwide>, <r>shrnb<mode>, <r>shrnt<mode>)
3448 (*aarch64_sve2_sra<mode>): Move further down file.
3449 * config/aarch64/t-aarch64 (s-check-sve-md): Check aarch64-sve2.md too.
3451 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3453 * config/aarch64/iterators.md (SVE_WHILE): Add UNSPEC_WHILERW
3455 (while_optab_cmp): Handle them.
3456 * config/aarch64/aarch64-sve.md
3457 (*while_<while_optab_cmp><GPI:mode><PRED_ALL:mode>_ptest): Make public
3458 and add a "@" marker.
3459 * config/aarch64/aarch64-sve2.md (check_<raw_war>_ptrs<mode>): Use it
3460 instead of gen_aarch64_sve2_while_ptest.
3461 (@aarch64_sve2_while<cmp_op><GPI:mode><PRED_ALL:mode>_ptest): Delete.
3463 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3465 * config/aarch64/aarch64.md (UNSPEC_WHILE_LE): Rename to...
3466 (UNSPEC_WHILELE): ...this.
3467 (UNSPEC_WHILE_LO): Rename to...
3468 (UNSPEC_WHILELO): ...this.
3469 (UNSPEC_WHILE_LS): Rename to...
3470 (UNSPEC_WHILELS): ...this.
3471 (UNSPEC_WHILE_LT): Rename to...
3472 (UNSPEC_WHILELT): ...this.
3473 * config/aarch64/iterators.md (SVE_WHILE): Update accordingly.
3474 (cmp_op, while_optab_cmp): Likewise.
3475 * config/aarch64/aarch64.c (aarch64_sve_move_pred_via_while): Likewise.
3476 * config/aarch64/aarch64-sve-builtins-base.cc (svwhilele): Likewise.
3477 (svwhilelt): Likewise.
3479 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3481 * config/aarch64/aarch64-sve-builtins-shapes.h (unary_count): Delete.
3482 (unary_to_uint): Define.
3483 * config/aarch64/aarch64-sve-builtins-shapes.cc (unary_count_def)
3484 (unary_count): Rename to...
3485 (unary_to_uint_def, unary_to_uint): ...this.
3486 * config/aarch64/aarch64-sve-builtins-base.def: Update accordingly.
3488 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3490 * config/aarch64/aarch64-sve-builtins-functions.h
3491 (code_for_mode_function): New class.
3492 (CODE_FOR_MODE0, QUIET_CODE_FOR_MODE0): New macros.
3493 * config/aarch64/aarch64-sve-builtins-base.cc (svcompact_impl)
3494 (svext_impl, svmul_lane_impl, svsplice_impl, svtmad_impl): Delete.
3495 (svcompact, svext, svsplice): Use QUIET_CODE_FOR_MODE0.
3496 (svmul_lane, svtmad): Use CODE_FOR_MODE0.
3498 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3500 * config/aarch64/iterators.md (addsub): New code attribute.
3501 * config/aarch64/aarch64-simd.md (aarch64_<su_optab><optab><mode>):
3503 (aarch64_<su_optab>q<addsub><mode>): ...this, making the same change
3504 in the asm string and attributes. Fix indentation.
3505 * config/aarch64/aarch64-sve.md (@aarch64_<su_optab><optab><mode>):
3507 (@aarch64_sve_<optab><mode>): ...this.
3508 * config/aarch64/aarch64-sve-builtins.h
3509 (function_expander::expand_signed_unpred_op): Delete.
3510 * config/aarch64/aarch64-sve-builtins.cc
3511 (function_expander::expand_signed_unpred_op): Likewise.
3512 (function_expander::map_to_rtx_codes): If the optab isn't defined,
3513 try using code_for_aarch64_sve instead.
3514 * config/aarch64/aarch64-sve-builtins-base.cc (svqadd_impl): Delete.
3515 (svqsub_impl): Likewise.
3516 (svqadd, svqsub): Use rtx_code_function instead.
3518 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3520 * config/aarch64/iterators.md (SRHSUB, URHSUB): Delete.
3521 (HADDSUB, sur, addsub): Remove them.
3523 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3525 * tree-nrv.c (pass_return_slot::execute): Handle all internal
3526 functions the same way, rather than singling out those that
3527 aren't mapped directly to optabs.
3529 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3531 * target.def (compatible_vector_types_p): New target hook.
3532 * hooks.h (hook_bool_const_tree_const_tree_true): Declare.
3533 * hooks.c (hook_bool_const_tree_const_tree_true): New function.
3534 * doc/tm.texi.in (TARGET_COMPATIBLE_VECTOR_TYPES_P): New hook.
3535 * doc/tm.texi: Regenerate.
3536 * gimple-expr.c: Include target.h.
3537 (useless_type_conversion_p): Use targetm.compatible_vector_types_p.
3538 * config/aarch64/aarch64.c (aarch64_compatible_vector_types_p): New
3540 (TARGET_COMPATIBLE_VECTOR_TYPES_P): Define.
3541 * config/aarch64/aarch64-sve-builtins.cc (gimple_folder::convert_pred):
3542 Use the original predicate if it already has a suitable type.
3544 2020-01-09 Martin Jambor <mjambor@suse.cz>
3546 * cgraph.h (cgraph_edge): Make remove, set_call_stmt, make_direct,
3547 resolve_speculation and redirect_call_stmt_to_callee static. Change
3548 return type of set_call_stmt to cgraph_edge *.
3549 * auto-profile.c (afdo_indirect_call): Adjust call to
3550 redirect_call_stmt_to_callee.
3551 * cgraph.c (cgraph_edge::set_call_stmt): Make return cgraph-edge *,
3552 make the this pointer explicit, adjust self-recursive calls and the
3553 call top make_direct. Return the resulting edge.
3554 (cgraph_edge::remove): Make this pointer explicit.
3555 (cgraph_edge::resolve_speculation): Likewise, adjust call to remove.
3556 (cgraph_edge::make_direct): Likewise, adjust call to
3557 resolve_speculation.
3558 (cgraph_edge::redirect_call_stmt_to_callee): Likewise, also adjust
3559 call to set_call_stmt.
3560 (cgraph_update_edges_for_call_stmt_node): Update call to
3561 set_call_stmt and remove.
3562 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
3563 Renamed edge to master_edge. Adjusted calls to set_call_stmt.
3564 (cgraph_node::create_edge_including_clones): Moved "first" definition
3565 of edge to the block where it was used. Adjusted calls to
3567 (cgraph_node::remove_symbol_and_inline_clones): Adjust call to
3568 cgraph_edge::remove.
3569 * cgraphunit.c (walk_polymorphic_call_targets): Adjusted calls to
3570 make_direct and redirect_call_stmt_to_callee.
3571 * ipa-fnsummary.c (redirect_to_unreachable): Adjust calls to
3572 resolve_speculation and make_direct.
3573 * ipa-inline-transform.c (inline_transform): Adjust call to
3574 redirect_call_stmt_to_callee.
3575 (check_speculations_1):: Adjust call to resolve_speculation.
3576 * ipa-inline.c (resolve_noninline_speculation): Adjust call to
3577 resolve-speculation.
3578 (inline_small_functions): Adjust call to resolve_speculation.
3579 (ipa_inline): Likewise.
3580 * ipa-prop.c (ipa_make_edge_direct_to_target): Adjust call to
3582 * ipa-visibility.c (function_and_variable_visibility): Make iteration
3583 safe with regards to edge removal, adjust calls to
3584 redirect_call_stmt_to_callee.
3585 * ipa.c (walk_polymorphic_call_targets): Adjust calls to make_direct
3586 and redirect_call_stmt_to_callee.
3587 * multiple_target.c (create_dispatcher_calls): Adjust call to
3588 redirect_call_stmt_to_callee
3589 (redirect_to_specific_clone): Likewise.
3590 * tree-cfgcleanup.c (delete_unreachable_blocks_update_callgraph):
3591 Adjust calls to cgraph_edge::remove.
3592 * tree-inline.c (copy_bb): Adjust call to set_call_stmt.
3593 (redirect_all_calls): Adjust call to redirect_call_stmt_to_callee.
3594 (expand_call_inline): Adjust call to cgraph_edge::remove.
3596 2020-01-09 Martin Liska <mliska@suse.cz>
3598 * params.opt: Set Optimization for
3599 param_max_speculative_devirt_maydefs.
3601 2020-01-09 Martin Sebor <msebor@redhat.com>
3605 * builtins.c (compute_objsize): Avoid handling MEM_REFs of vector type.
3607 2020-01-09 Martin Liska <mliska@suse.cz>
3609 * auto-profile.c (auto_profile): Use opt_for_fn
3611 * ipa-cp.c (ipcp_lattice::add_value): Likewise.
3612 (propagate_vals_across_arith_jfunc): Likewise.
3613 (hint_time_bonus): Likewise.
3614 (incorporate_penalties): Likewise.
3615 (good_cloning_opportunity_p): Likewise.
3616 (perform_estimation_of_a_value): Likewise.
3617 (estimate_local_effects): Likewise.
3618 (ipcp_propagate_stage): Likewise.
3619 * ipa-fnsummary.c (decompose_param_expr): Likewise.
3620 (set_switch_stmt_execution_predicate): Likewise.
3621 (analyze_function_body): Likewise.
3622 * ipa-inline-analysis.c (offline_size): Likewise.
3623 * ipa-inline.c (early_inliner): Likewise.
3624 * ipa-prop.c (ipa_analyze_node): Likewise.
3625 (ipcp_transform_function): Likewise.
3626 * ipa-sra.c (process_scan_results): Likewise.
3627 (ipa_sra_summarize_function): Likewise.
3628 * params.opt: Rename ipcp-unit-growth to
3629 ipa-cp-unit-growth. Add Optimization for various
3630 IPA-related parameters.
3632 2020-01-09 Richard Biener <rguenther@suse.de>
3635 * gimplify.c (gimplify_expr): Deal with NOP definitions.
3637 2020-01-09 Richard Biener <rguenther@suse.de>
3639 PR tree-optimization/93040
3640 * gimple-ssa-store-merging.c (find_bswap_or_nop): Raise search limit.
3642 2020-01-09 Georg-Johann Lay <avr@gjlay.de>
3644 * common/config/avr/avr-common.c (avr_option_optimization_table)
3645 [OPT_LEVELS_1_PLUS]: Set -fsplit-wide-types-early.
3647 2020-01-09 Martin Liska <mliska@suse.cz>
3649 * cgraphclones.c (symbol_table::materialize_all_clones):
3650 Use cgraph_node::dump_name.
3652 2020-01-09 Jakub Jelinek <jakub@redhat.com>
3655 * config/riscv/riscv.c (riscv_print_operand_reloc): Use
3656 output_operand_lossage instead of gcc_unreachable.
3657 * doc/md.texi (riscv f constraint): Fix typo.
3660 * config/i386/i386.md (subv<mode>4): Use SWIDWI iterator instead of
3661 SWI. Use <general_hilo_operand> instead of <general_operand>. Use
3662 CONST_SCALAR_INT_P instead of CONST_INT_P.
3663 (*subv<mode>4_1): Rename to ...
3664 (subv<mode>4_1): ... this.
3665 (*subv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
3666 define_insn_and_split patterns.
3667 (*subv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
3670 2020-01-08 David Malcolm <dmalcolm@redhat.com>
3672 * vec.c (class selftest::count_dtor): New class.
3673 (selftest::test_auto_delete_vec): New test.
3674 (selftest::vec_c_tests): Call it.
3675 * vec.h (class auto_delete_vec): New class template.
3676 (auto_delete_vec<T>::~auto_delete_vec): New dtor.
3678 2020-01-08 David Malcolm <dmalcolm@redhat.com>
3680 * sbitmap.h (auto_sbitmap): Add operator const_sbitmap.
3682 2020-01-08 Jim Wilson <jimw@sifive.com>
3684 * config/riscv/riscv.c (riscv_legitimize_tls_address): Ifdef out
3685 use of TLS_MODEL_LOCAL_EXEC when not pic.
3687 2020-01-08 David Malcolm <dmalcolm@redhat.com>
3689 * hash-map-tests.c (selftest::test_map_of_strings_to_int): Fix
3692 2020-01-08 Jakub Jelinek <jakub@redhat.com>
3695 * config/i386/i386.md (*stack_protect_set_2_<mode> peephole2,
3696 *stack_protect_set_3 peephole2): Also check that the second
3697 insns source is general_operand.
3700 * config/i386/i386.md (addcarry<mode>_0): Use nonimmediate_operand
3701 predicate for output operand instead of register_operand.
3702 (addcarry<mode>, addcarry<mode>_1): Likewise. Add alternative with
3703 memory destination and non-memory operands[2].
3705 2020-01-08 Martin Liska <mliska@suse.cz>
3707 * cgraph.c (cgraph_node::dump): Use ::dump_name or
3708 ::dump_asm_name instead of (::name or ::asm_name).
3709 * cgraphclones.c (symbol_table::materialize_all_clones): Likewise.
3710 * cgraphunit.c (walk_polymorphic_call_targets): Likewise.
3711 (analyze_functions): Likewise.
3712 (expand_all_functions): Likewise.
3713 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
3714 (propagate_bits_across_jump_function): Likewise.
3715 (dump_profile_updates): Likewise.
3716 (ipcp_store_bits_results): Likewise.
3717 (ipcp_store_vr_results): Likewise.
3718 * ipa-devirt.c (dump_targets): Likewise.
3719 * ipa-fnsummary.c (analyze_function_body): Likewise.
3720 * ipa-hsa.c (check_warn_node_versionable): Likewise.
3721 (process_hsa_functions): Likewise.
3722 * ipa-icf.c (sem_item_optimizer::merge_classes): Likewise.
3723 (set_alias_uids): Likewise.
3724 * ipa-inline-transform.c (save_inline_function_body): Likewise.
3725 * ipa-inline.c (recursive_inlining): Likewise.
3726 (inline_to_all_callers_1): Likewise.
3727 (ipa_inline): Likewise.
3728 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
3729 (ipa_propagate_frequency): Likewise.
3730 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
3731 (remove_described_reference): Likewise.
3732 * ipa-pure-const.c (worse_state): Likewise.
3733 (check_retval_uses): Likewise.
3734 (analyze_function): Likewise.
3735 (propagate_pure_const): Likewise.
3736 (propagate_nothrow): Likewise.
3737 (dump_malloc_lattice): Likewise.
3738 (propagate_malloc): Likewise.
3739 (pass_local_pure_const::execute): Likewise.
3740 * ipa-visibility.c (optimize_weakref): Likewise.
3741 (function_and_variable_visibility): Likewise.
3742 * ipa.c (symbol_table::remove_unreachable_nodes): Likewise.
3743 (ipa_discover_variable_flags): Likewise.
3744 * lto-streamer-out.c (output_function): Likewise.
3745 (output_constructor): Likewise.
3746 * tree-inline.c (copy_bb): Likewise.
3747 * tree-ssa-structalias.c (ipa_pta_execute): Likewise.
3748 * varpool.c (symbol_table::remove_unreferenced_decls): Likewise.
3750 2020-01-08 Richard Biener <rguenther@suse.de>
3753 * tree-eh.c (sink_clobbers): Update virtual operands for
3754 the first and last stmt only. Add a dry-run capability.
3755 (pass_lower_eh_dispatch::execute): Perform clobber sinking
3756 after CFG manipulations and in RPO order to catch all
3757 secondary opportunities reliably.
3759 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
3762 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
3764 2019-01-08 Richard Biener <rguenther@suse.de>
3767 * gimple-fold.c (rewrite_to_defined_overflow): Mark stmt modified.
3768 * tree-ssa-loop-im.c (move_computations_worker): Properly adjust
3769 virtual operand, also updating SSA use.
3770 * gimple-loop-interchange.cc (loop_cand::undo_simple_reduction):
3771 Update stmt after resetting virtual operand.
3772 (tree_loop_interchange::move_code_to_inner_loop): Likewise.
3773 * gimple-iterator.c (gsi_remove): When not removing the stmt
3774 permanently do not delink immediate uses or mark the stmt modified.
3776 2020-01-08 Martin Liska <mliska@suse.cz>
3778 * ipa-fnsummary.c (dump_ipa_call_summary): Use symtab_node::dump_name.
3779 (ipa_call_context::estimate_size_and_time): Likewise.
3780 (inline_analyze_function): Likewise.
3782 2020-01-08 Martin Liska <mliska@suse.cz>
3784 * cgraph.c (cgraph_node::dump): Use systematically
3787 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
3789 Add -nodevicespecs option for avr.
3792 * config/avr/avr.opt (-nodevicespecs): New driver option.
3793 * config/avr/driver-avr.c (avr_devicespecs_file): Only issue
3794 "-specs=device-specs/..." if that option is not set.
3795 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
3797 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
3799 Implement 64-bit double functions for avr.
3802 * config.gcc (tm_defines) [target=avr]: Support --with-libf7,
3803 --with-double-comparison.
3804 * doc/install.texi: Document them.
3805 * config/avr/avr-c.c (avr_cpu_cpp_builtins)
3806 <WITH_LIBF7_LIBGCC, WITH_LIBF7_MATH, WITH_LIBF7_MATH_SYMBOLS>
3807 <WITH_DOUBLE_COMPARISON>: New built-in defines.
3808 * doc/invoke.texi (AVR Built-in Macros): Document them.
3809 * config/avr/avr-protos.h (avr_float_lib_compare_returns_bool): New.
3810 * config/avr/avr.c (avr_float_lib_compare_returns_bool): New function.
3811 * config/avr/avr.h (FLOAT_LIB_COMPARE_RETURNS_BOOL): New macro.
3813 2020-01-08 Richard Earnshaw <rearnsha@arm.com>
3816 * config/arm/t-multilib (MULTILIB_MATCHES): Add rules to match
3817 armv7-a{+mp,+sec,+mp+sec} to appropriate armv7 multilib variants
3818 when only building rm-profile multilibs.
3820 2020-01-08 Feng Xue <fxue@os.amperecomputing.com>
3823 * ipa-cp.c (self_recursively_generated_p): Find matched aggregate
3824 lattice for a value to check.
3825 (propagate_vals_across_arith_jfunc): Add an assertion to ensure
3826 finite propagation in self-recursive scc.
3828 2020-01-08 Luo Xiong Hu <luoxhu@linux.ibm.com>
3830 * ipa-inline.c (caller_growth_limits): Restore the AND.
3832 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
3834 * config/gcn/gcn-valu.md (VEC_1REG_INT_ALT): Delete iterator.
3835 (VEC_ALLREG_ALT): New iterator.
3836 (VEC_ALLREG_INT_MODE): New iterator.
3837 (VCMP_MODE): New iterator.
3838 (VCMP_MODE_INT): New iterator.
3839 (vec_cmpu<mode>di): Use VCMP_MODE_INT.
3840 (vec_cmp<u>v64qidi): New define_expand.
3841 (vec_cmp<mode>di_exec): Use VCMP_MODE.
3842 (vec_cmpu<mode>di_exec): New define_expand.
3843 (vec_cmp<u>v64qidi_exec): New define_expand.
3844 (vec_cmp<mode>di_dup): Use VCMP_MODE.
3845 (vec_cmp<mode>di_dup_exec): Use VCMP_MODE.
3846 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>): Rename ...
3847 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): ... to this.
3848 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>_exec): Rename ...
3849 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): ... to this.
3850 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>): Rename ...
3851 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): ... to this.
3852 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>_exec): Rename ...
3853 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): ... to
3855 * config/gcn/gcn.c (print_operand): Fix 8 and 16 bit suffixes.
3856 * config/gcn/gcn.md (expander): Add sign_extend and zero_extend.
3858 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
3860 * config/gcn/constraints.md (DA): Update description and match.
3862 (Db): New constraint.
3863 * config/gcn/gcn-protos.h (gcn_inline_constant64_p): Add second
3865 * config/gcn/gcn.c (gcn_inline_constant64_p): Add 'mixed' parameter.
3866 Implement 'Db' mixed immediate type.
3867 * config/gcn/gcn-valu.md (addcv64si3<exec_vcc>): Rework constraints.
3868 (addcv64si3_dup<exec_vcc>): Delete.
3869 (subcv64si3<exec_vcc>): Rework constraints.
3870 (addv64di3): Rework constraints.
3871 (addv64di3_exec): Rework constraints.
3872 (subv64di3): Rework constraints.
3873 (addv64di3_dup): Delete.
3874 (addv64di3_dup_exec): Delete.
3875 (addv64di3_zext): Rework constraints.
3876 (addv64di3_zext_exec): Rework constraints.
3877 (addv64di3_zext_dup): Rework constraints.
3878 (addv64di3_zext_dup_exec): Rework constraints.
3879 (addv64di3_zext_dup2): Rework constraints.
3880 (addv64di3_zext_dup2_exec): Rework constraints.
3881 (addv64di3_sext_dup2): Rework constraints.
3882 (addv64di3_sext_dup2_exec): Rework constraints.
3884 2020-01-07 Andre Vieira <andre.simoesdiasvieira@arm.com>
3886 * doc/sourcebuild.texi (arm_little_endian, arm_nothumb): Documented
3887 existing target checks.
3889 2020-01-07 Richard Biener <rguenther@suse.de>
3891 * doc/install.texi: Bump minimal supported MPC version.
3893 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
3895 * langhooks-def.h (lhd_simulate_enum_decl): Declare.
3896 (LANG_HOOKS_SIMULATE_ENUM_DECL): Use it.
3897 * langhooks.c: Include stor-layout.h.
3898 (lhd_simulate_enum_decl): New function.
3899 * config/aarch64/aarch64-sve-builtins.cc (init_builtins): Call
3900 handle_arm_sve_h for the LTO frontend.
3901 (register_vector_type): Cope with null returns from pushdecl.
3903 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
3905 * config/aarch64/aarch64-protos.h (aarch64_sve::svbool_type_p)
3906 (aarch64_sve::nvectors_if_data_type): Replace with...
3907 (aarch64_sve::builtin_type_p): ...this.
3908 * config/aarch64/aarch64-sve-builtins.cc: Include attribs.h.
3909 (find_vector_type): Delete.
3910 (add_sve_type_attribute): New function.
3911 (lookup_sve_type_attribute): Likewise.
3912 (register_builtin_types): Add an "SVE type" attribute to each type.
3913 (register_tuple_type): Likewise.
3914 (svbool_type_p, nvectors_if_data_type): Delete.
3915 (mangle_builtin_type): Use lookup_sve_type_attribute.
3916 (builtin_type_p): Likewise. Add an overload that returns the
3917 number of constituent vector and predicate registers.
3918 * config/aarch64/aarch64.c (aarch64_sve_argument_p): Delete.
3919 (aarch64_returns_value_in_sve_regs_p): Use aarch64_sve::builtin_type_p
3920 instead of aarch64_sve_argument_p.
3921 (aarch64_takes_arguments_in_sve_regs_p): Likewise.
3922 (aarch64_pass_by_reference): Likewise.
3923 (aarch64_function_value_1): Likewise.
3924 (aarch64_return_in_memory): Likewise.
3925 (aarch64_layout_arg): Likewise.
3927 2020-01-07 Jakub Jelinek <jakub@redhat.com>
3929 PR tree-optimization/93156
3930 * tree-ssa-ccp.c (bit_value_binop): For x * x note that the second
3931 least significant bit is always clear.
3933 PR tree-optimization/93118
3934 * match.pd ((x >> c) << c -> x & (-1<<c)): Add nop_convert?. Add new
3935 simplifier with two intermediate conversions.
3937 2020-01-07 Martin Liska <mliska@suse.cz>
3939 * params.opt: Add Optimization for various parameters.
3941 2020-01-07 Martin Liska <mliska@suse.cz>
3944 * doc/extend.texi: Explain cloning for target_clone
3947 2020-01-07 Martin Liska <mliska@suse.cz>
3949 PR tree-optimization/92860
3950 * common.opt: Make in Optimization option
3951 as it is affected by -O0, which is an Optimization
3953 * tree-inline.c (tree_inlinable_function_p):
3954 Use opt_for_fn for warn_inline.
3955 (expand_call_inline): Likewise.
3957 2020-01-07 Martin Liska <mliska@suse.cz>
3959 PR tree-optimization/92860
3960 * common.opt: Make flag_ree as optimization
3963 2020-01-07 Martin Liska <mliska@suse.cz>
3965 PR optimization/92860
3966 * params.opt: Mark param_min_crossjump_insns with Optimization
3969 2020-01-07 Luo Xiong Hu <luoxhu@linux.ibm.com>
3971 * ipa-inline-analysis.c (estimate_growth): Fix typo.
3972 * ipa-inline.c (caller_growth_limits): Use OR instead of AND.
3974 2020-01-06 Michael Meissner <meissner@linux.ibm.com>
3976 * config/rs6000/rs6000.c (hard_reg_and_mode_to_addr_mask): New
3977 helper function to return the valid addressing formats for a given
3978 hard register and mode.
3979 (rs6000_adjust_vec_address): Call hard_reg_and_mode_to_addr_mask.
3981 * config/rs6000/constraints.md (Q constraint): Update
3983 * doc/md.texi (RS/6000 constraints): Update 'Q' cosntraint
3986 * config/rs6000/vsx.md (vsx_extract_<mode>_var, VSX_D iterator):
3987 Use 'Q' for doing vector extract from memory.
3988 (vsx_extract_v4sf_var): Use 'Q' for doing vector extract from
3990 (vsx_extract_<mode>_var, VSX_EXTRACT_I iterator): Use 'Q' for
3991 doing vector extract from memory.
3992 (vsx_extract_<mode>_<VS_scalar>mode_var): Use 'Q' for doing vector
3993 extract from memory.
3995 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add support
3996 for the offset being 34-bits when -mcpu=future is used.
3998 2020-01-06 John David Anglin <danglin@gcc.gnu.org>
4000 * config/pa/pa.md: Revert change to use ordered_comparison_operator
4001 instead of cmpib_comparison_operator in cmpib patterns.
4002 * config/pa/predicates.md (cmpib_comparison_operator): Revert removal
4003 of cmpib_comparison_operator. Revise comment.
4005 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4007 * tree-vect-slp.c (vect_build_slp_tree_1): Require all shifts
4008 in an IFN_DIV_POW2 node to be equal.
4010 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4012 * tree-vect-stmts.c (vect_check_load_store_mask): Rename to...
4013 (vect_check_scalar_mask): ...this.
4014 (vectorizable_store, vectorizable_load): Update call accordingly.
4015 (vectorizable_call): Use vect_check_scalar_mask to check the mask
4016 argument in calls to conditional internal functions.
4018 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
4020 * config/gcn/gcn-valu.md (subv64di3): Use separate alternatives for
4021 '0' matching inputs.
4022 (subv64di3_exec): Likewise.
4024 2020-01-06 Bryan Stenson <bryan@siliconvortex.com>
4026 * config/mips/mips.c (vr4130_align_insns): Fix typo.
4027 * doc/md.texi (movstr): Likewise.
4029 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
4031 * config/gcn/gcn-valu.md (vec_extract<mode><scalar_mode>): Add early
4034 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4036 * config/aarch64/t-aarch64 ($(srcdir)/config/aarch64/aarch64-tune.md):
4038 (s-aarch64-tune-md): ...this new stamp file. Pipe the new contents
4039 to a temporary file and use move-if-change to update the real
4040 file where necessary.
4042 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4044 * config/aarch64/aarch64-sve.md (@aarch64_sel_dup<mode>): Use Upl
4045 rather than Upa for CPY /M.
4047 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
4049 * config/gcn/gcn.c (gcn_inline_constant_p): Allow 64 as an inline
4052 2020-01-06 Martin Liska <mliska@suse.cz>
4054 PR tree-optimization/92860
4055 * params.opt: Mark param_max_combine_insns with Optimization
4058 2020-01-05 Jakub Jelinek <jakub@redhat.com>
4061 * config/i386/i386.md (SWIDWI): New mode iterator.
4062 (DWI, dwi): Add TImode variants.
4063 (addv<mode>4): Use SWIDWI iterator instead of SWI. Use
4064 <general_hilo_operand> instead of <general_operand>. Use
4065 CONST_SCALAR_INT_P instead of CONST_INT_P.
4066 (*addv<mode>4_1): Rename to ...
4067 (addv<mode>4_1): ... this.
4068 (QWI): New mode attribute.
4069 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
4070 define_insn_and_split patterns.
4071 (*addv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
4073 (uaddv<mode>4): Use SWIDWI iterator instead of SWI. Use
4074 <general_hilo_operand> instead of <general_operand>.
4075 (*addcarry<mode>_1): New define_insn.
4076 (*add<dwi>3_doubleword_cc_overflow_1): New define_insn_and_split.
4078 2020-01-03 Konstantin Kharlamov <Hi-Angel@yandex.ru>
4080 * gdbinit.in (pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, pdd, pbs, pbm):
4081 Use "call" instead of "set".
4083 2020-01-03 Martin Jambor <mjambor@suse.cz>
4086 * ipa-cp.c (print_all_lattices): Skip functions without info.
4088 2020-01-03 Jakub Jelinek <jakub@redhat.com>
4091 * config/i386/i386-options.c (ix86_simd_clone_adjust): If
4092 TARGET_PREFER_AVX128, use prefer-vector-width=256 for 'c' and 'd'
4093 simd clones. If TARGET_PREFER_AVX256, use prefer-vector-width=512
4094 for 'e' simd clones.
4097 * config/i386/i386.opt (x_prefer_vector_width_type): Remove TargetSave
4099 (mprefer-vector-width=): Add Save.
4100 * config/i386/i386-options.c (ix86_target_string): Add PVW argument, print
4101 -mprefer-vector-width= if non-zero. Fix up -mfpmath= comment.
4102 (ix86_debug_options, ix86_function_specific_print): Adjust
4103 ix86_target_string callers.
4104 (ix86_valid_target_attribute_inner_p): Handle prefer-vector-width=.
4105 (ix86_valid_target_attribute_tree): Likewise.
4106 * config/i386/i386-options.h (ix86_target_string): Add PVW argument.
4107 * config/i386/i386-expand.c (ix86_expand_builtin): Adjust
4108 ix86_target_string caller.
4111 * config/i386/i386.md (abs<mode>2): Use expand_simple_binop instead of
4112 emitting ASHIFTRT, XOR and MINUS by hand. Use gen_int_mode with QImode
4113 instead of gen_int_shift_amount + convert_modes.
4115 PR rtl-optimization/93088
4116 * loop-iv.c (find_single_def_src): Punt after looking through
4117 128 reg copies for regs with single definitions. Move definitions
4120 2020-01-02 Dennis Zhang <dennis.zhang@arm.com>
4122 * config/arm/arm-c.c (arm_cpu_builtins): Define
4123 __ARM_FEATURE_MATMUL_INT8, __ARM_FEATURE_BF16_VECTOR_ARITHMETIC,
4124 __ARM_FEATURE_BF16_SCALAR_ARITHMETIC, and
4125 __ARM_BF16_FORMAT_ALTERNATIVE when enabled.
4126 * config/arm/arm-cpus.in (armv8_6, i8mm, bf16): New features.
4127 * config/arm/arm-tables.opt: Regenerated.
4128 * config/arm/arm.c (arm_option_reconfigure_globals): Initialize
4129 arm_arch_i8mm and arm_arch_bf16 when enabled.
4130 * config/arm/arm.h (TARGET_I8MM): New macro.
4131 (TARGET_BF16_FP, TARGET_BF16_SIMD): Likewise.
4132 * config/arm/t-aprofile: Add matching rules for -march=armv8.6-a.
4133 * config/arm/t-arm-elf (all_v8_archs): Add armv8.6-a.
4134 * config/arm/t-multilib: Add matching rules for -march=armv8.6-a.
4135 (v8_6_a_simd_variants): New.
4136 (v8_*_a_simd_variants): Add i8mm and bf16.
4137 * doc/invoke.texi (armv8.6-a, i8mm, bf16): Document new options.
4139 2020-01-02 Jakub Jelinek <jakub@redhat.com>
4142 * predict.c (compute_function_frequency): Don't call
4143 warn_function_cold on functions that already have cold attribute.
4145 2020-01-01 John David Anglin <danglin@gcc.gnu.org>
4148 * config/pa/pa.c (pa_elf_select_rtx_section): New. Put references to
4149 COMDAT group function labels in .data.rel.ro.local section.
4150 * config/pa/pa32-linux.h (TARGET_ASM_SELECT_RTX_SECTION): Define.
4153 * config/pa/pa.md (scc): Use ordered_comparison_operator instead of
4154 comparison_operator in B and S integer comparisons. Likewise, use
4155 ordered_comparison_operator instead of cmpib_comparison_operator in
4157 * config/pa/predicates.md (cmpib_comparison_operator): Remove.
4159 2020-01-01 Jakub Jelinek <jakub@redhat.com>
4161 Update copyright years.
4163 * gcc.c (process_command): Update copyright notice dates.
4164 * gcov-dump.c (print_version): Ditto.
4165 * gcov.c (print_version): Ditto.
4166 * gcov-tool.c (print_version): Ditto.
4167 * gengtype.c (create_file): Ditto.
4168 * doc/cpp.texi: Bump @copying's copyright year.
4169 * doc/cppinternals.texi: Ditto.
4170 * doc/gcc.texi: Ditto.
4171 * doc/gccint.texi: Ditto.
4172 * doc/gcov.texi: Ditto.
4173 * doc/install.texi: Ditto.
4174 * doc/invoke.texi: Ditto.
4176 2020-01-01 Jan Hubicka <hubicka@ucw.cz>
4178 * ipa.c (walk_polymorphic_call_targets): Fix updating of overall
4181 2020-01-01 Jakub Jelinek <jakub@redhat.com>
4183 PR tree-optimization/93098
4184 * match.pd (popcount): For shift amounts, use integer_onep
4185 or wi::to_widest () == cst instead of tree_to_uhwi () == cst
4186 tests. Make sure that precision is power of two larger than or equal
4187 to 16. Ensure shift is never negative. Use HOST_WIDE_INT_UC macro
4188 instead of ULL suffixed constants. Formatting fixes.
4190 Copyright (C) 2020 Free Software Foundation, Inc.
4192 Copying and distribution of this file, with or without modification,
4193 are permitted in any medium without royalty provided the copyright
4194 notice and this notice are preserved.